Jump to text
On this page
Copy and source
Published source confirmed at last check

Source changed 2026-11-16 21:18:08 UTC · snapshot created 2026-11-16 21:19:40 UTC · last check 2026-11-17 19:15:10 UTC

A. Formulate the subject question

MMP.10 - Construct and Revise a Mathematical Constraint Formulation

Type: Method pattern Status: Stable Normativity: Normative within the stated use

MMP.10:1 - Problem frame

Use this pattern when you know the conditions a possible object or situation must satisfy and need to express its possibilities mathematically. You may be looking for an arrangement, a quantity, a shape or a rule of action. Representations such as a number for an amount, an ordered list, a set of members and an unknown function support different operations and possible answers.

Start by describing one candidate object and what would make it admissible. Choose a way to represent it, then distinguish two kinds of condition: those needed for the representation to denote such an object, and those expressing the requirements on that object. This produces a mathematical problem that can be reasoned about or handed to an appropriate computational method.

The result consists of variables with their domains, joint conditions, the requested operation on their solutions and a way to interpret the answer. It can be enough to find one feasible case; another question may require all possible values, a preferred case or a count. The chosen representation must support the requested result.

You need to understand the objects and requirements in the original question, elementary logical conditions and the mathematics used to express them. A mathematical collaborator can construct the expressions while you resolve their intended meaning. When a suitable formulation already exists, use it. When the missing contribution is a physical relation, an observation law or a rule of change, obtain that relation through the relevant method before translating it. B.5.FM supplies the broader first-model construction; MMP.7 and MMP.9 develop observation and reduced-evolution constructions.

MMP.10:2 - Problem

A convenient variable can leave out the possibility that matters. A single interval cannot represent an activity that may pause. A membership bit cannot record repeated membership. Coefficients of a straight line cannot describe every continuous curve. These restrictions may be useful, but they change which cases the formulation can answer for.

Representation can also add possibilities. A table intended to describe a function can allow two values for one argument unless its conditions exclude that case. Extra fields can create several records of the same object. A calculation over those records can then answer a different counting or probability question.

The difficulty is to construct the mathematical expression of the intended possibilities, including their joint restrictions and the requested result. Solving the resulting equations addresses the problem only after that construction has been made.

MMP.10:3 - Forces

ForceConsequence for the construction
Faithful possibilities and useful operationsA familiar representation may make calculation easy while excluding an intended case or hiding a useful relation.
Structured objects and scalar toolsA function, set or sequence may need several scalar variables and additional conditions to represent its structure.
Shared restrictions and local descriptionsSeparate bounds on individual variables can lose a condition on their combination.
Simple answer and rich solution setA single witness needs less from the representation than counting, sampling or claiming that no witness exists.
Reuse and changed requirementsA derived constraint that helped an earlier formulation may exclude valid cases after the requirement changes.

MMP.10:4 - Solution

Describe the possible object → choose its representation → derive the representation conditions → express the requirements and question → obtain and interpret a result → revise the affected construction.

MMP.10:4.1 - Recover what varies and what the question asks

Describe a candidate before choosing scalar variables. Is it an amount, a collection, a sequence, an assignment, a function or another mathematical object? Which distinctions can change the answer? A set retains membership; a multiset also retains multiplicity; a sequence retains positions. Choose among them from the question.

Separate supplied quantities from unknowns. Among unknowns, distinguish what is to be inferred, what can be chosen and what can vary independently of that choice. If a choice uses an observation, specify when the observation becomes available; MMP.8 constructs the resulting information-dependent requirement. An unknown value does not become a freely selectable design variable merely by appearing in the same equation.

State the requested result. Existence asks whether at least one admissible object can be constructed. Inference asks what a quantity can be across admissible objects. Selection adds a preference among them. Counting and sampling depend on how individual objects are distinguished. Keep those requests separate while choosing the representation.

MMP.10:4.2 - Construct variables that represent the object

Choose an expression from which a candidate can be recovered. Give each variable a domain and any unit or reference point needed by its operations. A machine label ranges over names; arithmetic on the label requires a separate meaning. A count ranges over nonnegative integers; an amount may be divisible. State a finite bound when the task supplies one. Adding a bound solely to finish a search restricts the question to that bound.

For a structured object, compare representations by the operations you need. A function on a finite set can use one output variable for each input. Alternatively, a table of Boolean indicators can say which input-output pairs belong to its graph. The first expression makes function evaluation easy to state. The second makes some relations among pairs visible, but needs conditions to make the table a function. For a partial function, represent undefinedness as well as defined values.

Keep a shared quantity shared. If several equations use the same unknown offset, one offset variable must occur in all of them. Introducing a separate offset in each equation creates additional possibilities. Conversely, equating values that can differ can remove possibilities.

For a function or shape over an infinite domain, choosing finitely many coefficients also chooses a family. Identify that family and whether it expresses the intended possibilities or is a deliberate restriction. For example, the conditions on a continuous function may allow curved solutions even when no affine function satisfies them. A useful restricted family can be sufficient for finding a witness; failure inside it leaves the larger family unresolved.

MMP.10:4.3 - Derive structural conditions and translate requirements

Ask what must hold for a variable assignment to describe one candidate of the intended kind. With Boolean entries r_ij describing the graph of a total function, require sum_j r_ij = 1 for every input i. For a partial function, replace this with sum_j r_ij <= 1; an all-zero row then means undefined at that input. For an injective function, additional conditions on columns express the extra requirement.

These conditions have different reasons. One value per input comes from choosing a total function. Add the column conditions when the problem requires injectivity. Keep their reasons recoverable so that a later change from total to partial or from injective to unrestricted has a local repair.

Express the original requirements using the represented objects. Conjoin conditions that must hold for the same assignment. Use disjunction for allowed alternatives and implication when choosing an option imposes a condition. An implication alone supplies no timing; use time quantities or an explicit sequence when order matters. Preserve a coupled condition such as x+y=1 rather than replacing it by separate bounds on x and y.

When a bijection between representations is established, MATH.7 supplies transport of operations, relations and compound expressions. For a representation with several records per object, C.29.1 supplies the more general correspondence. Use the decoding of a record to express the requirement on its object. If a condition is rewritten to fit the receiving notation, derive that expression from the original relation and the structural conditions. This is where a missing index, an undefined value or a lost alternative can change the formulation.

Additional constraints can expose consequences and help the obtaining method. Derive them from the retained requirements, and preserve that dependence. When efficiency matters, compare the work each formulation requires. A formulation with fewer variables or more constraints can still take longer to solve.

MMP.10:4.4 - Make the answer correspond to the question

Define how to recover the requested object or quantity from a satisfying assignment. Then work in both directions: represent an intended admissible case, and interpret an allowed assignment. Use the construction and its conditions to establish the reach of this correspondence. A small case can expose a mistake; a claim about every case needs the corresponding argument.

Match that reach to the requested result:

  • To use a witness, its recovered object must satisfy the original requirements.
  • To conclude that no intended object exists from inconsistency of the formulation, every intended object must have a representation in it.
  • To infer all possible values, translate the quantity as well as the admissible cases; C.16.IR supplies the projection question.
  • To optimize, translate the objective and preference as well as feasibility. Distinguish a bound from a value attained by an object.
  • To count or sample objects, account for multiple representations of the same object. One representation per object is one solution; weighting or grouping representations can be another.

Auxiliary variables can change what a returned number means. Suppose a finite nonempty set of finish times f_i is determined by the other variables. Introduce a real auxiliary T used only in T >= f_i and the objective of minimizing T. Lowering T to max_i f_i then preserves feasibility, so at an attained optimum T equals the latest finish. If T must instead be an integer and the latest finish is 1/2, its minimum is 1. A merely feasible intermediate T can also exceed the latest finish. Recover the latest finish as max_i f_i; infer equality with T only when its domain and other conditions permit that lowering.

Choose an obtaining method for the constructed question. Manual substitution may suffice; another problem needs a numerical method, symbolic derivation or search. C.29.2 separates the required mathematical result from the procedure and its execution. Use the conclusion the method supports: finding no case within a time budget differs from establishing inconsistency. Preserve any restriction or approximation when returning the result to the original question through C.29.1.

MMP.10:4.5 - Revise the formulation from the changed requirement

Locate the changed participant, domain, relation or requested result. Changing a supplied amount can retain the same representation. Allowing interruptions changes what an activity description must express. Changing a total function to a partial one changes structural conditions. Changing existence to counting can make duplicate records material.

Revisit constraints derived from the old requirement as well as the original formula. Reconstruct the affected expressions and answer interpretation, retaining the unaffected ones. If a solving tool cannot support the needed object, construct a suitable representation or choose another obtaining method; keep any deliberate restriction visible in the returned conclusion.

Stop with a usable formulation and interpretation, an adequate answer, or a named missing relation or operation. Choose further derivation, observation or computation according to what it can change in the work and its cost; C.11.DUA supplies that decision. When the formulated object is itself a working method, use MMP.22 when the mathematical transformation still needs to be reconstructed as a working rule, then return the candidate to C.32.MWA.PP for practical comparison and A.3.1.MV where it changes a reusable Method. The mathematical model supplies a reason for the change; the working method still has to be performed under its stated conditions.

MMP.10:5 - Archetypal Grounding

MMP.10:5.1 - Construct an unknown rule from requirements on its repetitions

A device has three labeled modes A, B and C. The required rule changes the mode on every use and returns to the starting mode after three uses. The question is to construct a deterministic rule, with no additional internal state. The rule itself is the unknown object.

Let S={A,B,C}. Choose one output variable p_i in S for each input i. The requirements become p_i != i and p_(p_(p_i)) = i for every i. Function composition gives the meaning of the repeated application. MATH.1 constructs composable paths; MATH.5 extends an interpretation of their elementary steps to the compounds.

To express the rule by selected pairs instead, choose r_ij in {0,1}. Add sum_j r_ij=1 for each row, r_ii=0, and, for all i,j,k, (r_ij=1 AND r_jk=1) implies r_ki=1. The row condition makes a function. The implication expresses the return after three uses: the first two selected transitions determine a required third.

Recover p by taking the unique selected column in each row. Conversely, p creates the table by selecting exactly its output pair in each row. These constructions are inverse. The triple-application requirement is therefore the same in both formulations, with MATH.7 carrying that relation. A rule A→B→C→A and its reverse both satisfy it.

There are precisely two such rules. From p^3=id, p is invertible with inverse p^2. Its cycles have lengths dividing three. Since a one-element cycle is forbidden, the three modes form one three-element cycle, with two possible orientations. The cycle argument rules out any further solutions.

Change the device to four modes, retaining the three-use return and no unchanged mode. A permutation of four elements cannot partition them into cycles all of length three, so no rule exists under these conditions. Change instead to a return after two uses. The conditions become p_(p_i)=i and p_i!=i; the indicator formulation requires symmetry r_ij=r_ji. It admits three pairings of four modes. Remove the former three-use implication: leaving it in the formulation would make the new, feasible requirement appear impossible.

The result is a rule that can be implemented and its stated scope: deterministic changes of the visible mode without hidden state. A proposal with additional state describes a different device and needs a new account of its operation.

MMP.10:5.2 - Preserve existence while repairing a count

An optional assignment gives each of two named requests either no selected option or one of options 0 and 1. Different requests may select the same option. This is a partial function from the two requests to {0,1}. Each request has three possibilities, so there are nine assignments.

Suppose the storage format gives each request two bits: d says whether an option is defined, and q gives its value when defined. When d=0, q is ignored. All sixteen four-bit records denote valid partial assignments. Every assignment has a record, so the representation can support an existence query with translated requirements.

It does not preserve the count. The empty assignment has four records, each of the four assignments defined on exactly one request has two records, and each of the four total assignments has one record. Thus 4 + 4*2 + 4 = 16. Uniform selection among records gives probability 4/16 to the empty assignment and 1/16 to each total assignment, rather than the 1/9 obtained by uniform selection among assignments.

For a count or uniform assignment sample, one repair is the structural condition d=0 implies q=0 for each request. There are now three admissible bit pairs per request and nine records, one per assignment. Another is a single variable with domain {absent,0,1} per request. If the sixteen-record storage representation must remain, group or weight its records using the multiplicities instead. To sample the nine assignments uniformly, give each record of an assignment with m records probability 1/(9*m). The probabilities of all m records then sum to 1/9 for that assignment. Thus each record of the empty assignment receives 1/36, each record of a one-request assignment 1/18, and each total-assignment record 1/9. C.29.1 supplies the required correspondence; MMP.7 supplies a probability law when sampling is the intended operation.

The sixteen-record representation still supports the original existence query. For a count or a uniform sample of the same nine assignments, account for their different numbers of records.

MMP.10:5.3 - Keep quantities and domain restrictions together

A preparation requires one litre containing 35 percent solute by volume, using solutions A and B at 20 and 80 percent. Assume solute is conserved and component volumes add in this preparation. Those subject assumptions supply the relations. Let x and y be the respective volumes in litres; choose nonnegative real domains because the amounts can initially be divided freely.

The joint conditions are x+y=1 and 0.2*x+0.8*y=0.35. Substitution gives x=0.75, y=0.25. Both the total and solute requirements hold for those amounts. Separate bounds 0<=x<=1 and 0<=y<=1 would lose their required total.

Now only whole half-litre doses may be used. Change the representation to x=m/2, y=n/2, with nonnegative integers m and n. The volume equation becomes m+n=2. Its possibilities (m,n)=(2,0),(1,1),(0,2) give solute amounts 0.2, 0.5 and 0.8 litre. None supplies 0.35 litre. Rounding the former solution changes the preparation; it does not satisfy its original condition.

If the required concentration changes to 50 percent, one dose of each solution works. The subject relations and unit remain, while the requirement and feasible assignment change. If mixing changes volume or solute, obtain the replacement subject relation before revising its mathematical expression.

MMP.10:6 - Bias-Annotation

Tool familiarity can make scalar variables appear inevitable and conceal a different object or useful operation. Begin with the candidate object and its requirements. Compare representations when that comparison can change the answer or obtaining effort.

Compactness can hide duplicate records or omitted possibilities. Judge the representation by the requested result and the work needed to obtain it.

MMP.10:7 - Conformance Checklist

  • Can the reader identify a candidate object before interpreting the variables, including what is supplied, unknown or selectable?
  • Do domains, units, shared quantities and definedness express the intended possibilities?
  • Which conditions make the representation denote an object, and which express requirements on it?
  • Do conditions that must hold together refer to the same assignment? Are alternatives and conditional restrictions preserved?
  • Can intended cases be represented and satisfying assignments be interpreted at the reach required by the question?
  • Does the requested value, preference, count or probability survive the representation, including auxiliary values and duplicate records?
  • What conclusion does the obtaining method establish, and which changed requirement reopens which construction?

MMP.10:8 - Common Anti-Patterns and How to Avoid Them

FailureWhy it changes the resultRepair
Use separate bounds in place of a joint relation.The same assignment can violate the lost total or coupling.Retain the relation with its shared variables.
Number labels and use their arithmetic as a subject law.Addition or order on the labels may describe no operation on the labeled objects.Supply the intended operation or use names as names.
Read an empty restricted search as absence in the original problem.Intended cases may lie outside the chosen bound or family.Return the restricted conclusion or extend the representation.
Count records as objects after adding auxiliary fields.Multiple records can describe one object and alter the count or sampling law.Canonicalize, group or weight by the needed correspondence.
Keep a consequence of a replaced requirement.It can remove the newly allowed cases, as in the changed rule in :5.1.Re-derive the affected constraints.

MMP.10:9 - Consequences

The mathematical problem becomes available for reasoning, computation and revision without leaving the interpretation of its variables implicit. A solver result can be returned as the arrangement, quantity or unresolved distinction that the work needs.

Constructing and maintaining a second representation has a cost. Direct formulation is often sufficient when the objects and their restrictions already have a clear expression. A structured intermediate formulation becomes useful when it preserves meaning across several receiving notations, helps revise requirements or exposes a calculation that the first representation hid. Any efficiency advantage depends on the resulting obtaining method.

MMP.10:10 - Architectural Rationale

Mathematical modeling often starts with choosing how possible objects will be expressed. The choice determines which restrictions must be added and which results can be recovered. Separating representation conditions, subject requirements and the operation on solutions makes revisions local: a changed requirement need not replace the representation, and a new representation need not change the intended possibilities.

An operation can itself be the unknown object. The finite-rule case therefore treats repeated action as a requirement on a function and uses composition to express it. The same organization applies to other structured objects, while their mathematics supplies the needed constructions and proofs. MATH.7 explains reversible transport after the maps are available; this pattern develops the modeling choice and construction of those expressions, including cases that need a many-to-one correspondence.

The connection to computation runs in both directions. A mathematical formulation supplies the problem a procedure must answer. The operations supported by a procedure can suggest a different expression of that problem. Meaning is retained through the representation conditions and answer interpretation, while performance is judged on the resulting work.

MMP.10:11 - SoTA-Echoing

The MiniZinc Handbook 2.10.1, modeling and efficiency sections develops alternative models, derived constraints and interactions with solving methods. Compare the work required to solve the alternative formulations, including when one has fewer variables. A direct scalar formulation remains economical when its meaning is already clear.

Akgun and colleagues, Conjure (2023), sections 2-4 separate representation selection from expression refinement and introduce structural constraints during refinement. Adopt this construction when structured objects would otherwise disappear into unexplained scalar choices. Conjure’s finite combinatorial scope and model-selection heuristic remain specific to that approach.

The Essence language reference, function and relation domains makes such choices as partiality and cardinality explicit. Use those distinctions in :4.2–4.3. The same conditions can be stated directly in a familiar mathematical notation.

For infinitely many possible objects, a finite parameterization needs its own coverage or approximation argument. Consider continuous nonnegative functions f on [0,1], with f(0)=f(1)=0 and integral one. An affine parameterization permits only the zero function after the endpoint conditions, so it fails. The quadratic f(t)=6*t*(1-t) satisfies every requirement. The quadratic provides a witness that the affine family omits. C.29.1 supplies the interpretation of a restriction; the relevant mathematical method supplies a suitable larger family or approximation.

To apply the formulation in another subject, obtain the relations and mathematical operations needed to express its requirements. Reconsider the representation when a new requirement, result kind or obtaining method changes what it needs to preserve or make affordable.

MMP.10:12 - Relations

MMP.21 uses the formulation to compare connected operations, preserved observations and shared capacity. MMP.22 uses a mathematical transformation to construct the changed reusable working rule and its necessary conditions.

  • B.5.FM and B.5.TU: construct a first account and connect a subject theory to the encountered problem. This pattern translates its candidate objects and conditions into a mathematical formulation.
  • MATH.1 and MATH.5: construct composable paths and extend an assignment on elementary steps to compounds while preserving operations and equations. MATH.16 chooses a mathematical construction from the maps it must support. MATH.7 transports structure when the required bijections have been constructed.
  • C.29.1 and C.29.2: supply correspondence, answer recovery and the separation of a mathematical result from its obtaining procedure and execution.
  • C.16.IR: determines what a compatible set permits one to infer through projection and constancy of the requested quantity.
  • MMP.7 and MMP.8: construct observation probabilities and information-dependent choices. A formulation’s variables retain those probabilistic and temporal meanings.
  • MMP.9: derives a reduced evolution law when retained quantities depend on eliminated contributions. The resulting law can supply relations used here.
  • C.11.DUA, E.22 and E.23: choose worthwhile further inquiry and organize evaluation and improvement of a formulation or its obtaining work.
  • MMP.22 constructs a working-rule candidate from a mathematical transformation; C.32.MWA.PP compares practical arrangements and A.3.1.MV maintains a consequential Method variant. Its performance in the subject remains distinct from the formal properties of its description.

MMP.10:End

MMP.11 - Construct a Mathematical Model Family from Known Relations

Type: Method pattern Status: Stable Normativity: Normative within the stated use

MMP.11:1 - Problem frame

Use this pattern when part of a model is supported, but a function or other relation is still unknown. You may know which quantities interact, what must be conserved, or how a response begins and ends, while lacking its form between those conditions. You need candidate models that retain this knowledge while allowing the missing contribution to vary.

Begin by locating the unknown contribution in the relations needed for the question. Say what it takes as input, what it supplies and which properties its variation must preserve. Construct a family for that contribution, then put it back into the model. For a response that rises from zero to one, a family of curves constrained by those properties gives a different starting point from an unrestricted fitted curve.

The first useful result is a family of models, its remaining adjustable parts and the restrictions introduced by its construction. It may already give a sufficient bound. When observations are available, the same construction lets you ask which functions, parameters or consequences those observations can distinguish.

You need the subject grounds for the retained relations and enough mathematics to construct and use the family, or access to that mathematical contribution. You can use a known adequate relation directly. An unknown numerical parameter inside a suitable family normally needs its estimation method. Use this pattern when the family itself needs construction or revision. B.5.FM and B.5.TU help when the missing contribution is the subject account or theory from which a relation should come.

MMP.11:2 - Problem

A familiar formula can exclude the response being sought. Giving a learner unrestricted freedom can create the opposite problem: the fitted response may violate a known relation. Fitting two interacting contributions independently can also destroy a property that depends on their connection.

Agreement with recorded outputs leaves another difficulty. Several parameter settings may define the same function, and several functions may produce the same observations. Some receiving questions distinguish those alternatives; others need only a consequence on which they agree. Choosing one fitted instance can conceal this difference.

The modeling work is to construct the adjustable contribution in a form that retains the supported relations, then determine what that family can establish for the present question. The construction itself may impose additional restrictions, so its expressive limits belong to the answer.

MMP.11:3 - Forces

ForceTension
Retained knowledge and flexibilityA structural relation can rule out impossible candidates; an unsupported restriction can remove the needed one.
Local fitting and coupled behaviorA contribution can fit its own samples while disrupting the model in which it is used.
Simple representation and family coverageA small parameterization is easier to fit but can omit admissible functions.
Parameter recovery and useful inferenceParameters may remain ambiguous while a required consequence is determined.
Prediction and interventionTwo accounts can agree during observation and differ after one mechanism is changed.

MMP.11:4 - Solution

Locate the missing relation, separate what is retained from what may vary, construct the adjustable family, and derive its contribution to the receiving question. Observation or computation is then selected for what remains unresolved.

MMP.11:4.1 - Locate the contribution that may change

State the result wanted from the model. Identify the unknown relation and the quantities it connects. Keep its inputs, output, domain, units and permitted dependence explicit enough to substitute a candidate into the surrounding relations. A function of present state, a function of its history and a random response law admit different constructions. If equal proposed inputs require different deterministic outputs in the admitted circumstances, revise the inputs or retain those alternatives. Greater flexibility of a single-valued function cannot supply both outputs. For a dynamic model, A.3.3.TR supplies the corresponding reconsideration of state.

Recover the grounds and application range of the relations you retain. A balance may be required by the chosen boundary; monotonicity may hold only over one operating range; a shape assumption may be provisional. Keep an allowed discrepancy when the subject account supplies one.

Use the smallest part that can be varied without silently changing another retained claim. If an adjustable term can absorb a known contribution, include that possibility in the inference question. Section :5.2 shows an ambiguity between two gross transfers even when their net effect is known.

MMP.11:4.2 - Construct the permitted variation

Translate each retained property into a mathematical condition, then choose a construction that satisfies it. MMP.10 supplies the general work of representing candidate objects and their conditions. Here the object being constructed is a family of relations to insert into the model. Direct conditions may give a workable representation of that family. Constructing through free elements is useful when their variation should preserve the conditions; compare its obtaining and revision operations with those of the direct representation.

A useful construction separates a fixed part from free variation. If the unknown g must satisfy L(g)=b for a given linear operator L, find one particular solution g0 and choose a correction h with L(h)=0. Then g=g0+h retains the condition. This describes every solution only if the admitted corrections cover the whole null space in the chosen function domain. Restricting h to a few basis functions supplies a smaller family. Establish the linearity and domain before using this construction.

For a sign, bound or shape condition, construct through a map whose output has that property. Nonnegative weights can be normalized to probabilities. Integrating a nonnegative function can produce a nondecreasing response. Work out the domain and boundary of the resulting family: strict positivity excludes zeros, and an integral of an ordinary integrable function produces an absolutely continuous curve. Section :5.1 develops one such construction and its restriction.

When a property depends on coupling, construct the coupled contribution. Using the same transfer with opposite signs in two balance equations preserves their total. Two separately fitted right-hand sides have no such identity unless their joint conditions supply it. Use A.3.3.TR to assemble a change rule from the interacting relations.

A penalty during fitting offers a different construction: it discourages violations while allowing them. Use it when that allowance fits the question. If the account requires an identity, either build it into the representation or use an obtaining method that enforces it. The size of a training penalty does not by itself establish the identity.

MMP.11:4.3 - Choose the representation and its range

Choose a representation whose operations fit the required use and available resources. A table can represent a finite function. A basis expansion or program can retain a useful structure. A neural representation can supply a flexible adjustable function. The meaning of its inputs and outputs, and the retained relations, remain part of the model.

Check two different questions. Does every admitted parameter setting produce a relation allowed by the construction? Does the construction cover all relations needed for the present conclusion? A witness may need only one candidate; an impossibility claim over all admissible models needs coverage of that whole family or another sufficient argument.

State restrictions introduced by knots, basis functions, regularity, network architecture or domain truncation when they can change the answer. A numerical fit inside the restricted family answers for that family. If its consequence is sufficient, a more flexible family may add only cost. If the missing case matters, change the representation.

Different parameters can denote the same function. Recover the function or consequence needed by the receiving use. Distinguish parameters when the receiving question depends on them. Section :5.3 gives a normalization redundancy. Use MATH.7 when a change of representation has constructed inverse maps; use C.29.1 when correspondence is more general.

MMP.11:4.4 - Put the family into the model before using its fit

Substitute the adjustable contribution into the relations that consume it. Derive the resulting observable or answer condition with shared quantities kept shared. An error measured on an isolated contribution and an error in the coupled output are different fitting questions.

Choose the insertion point from what must remain meaningful. In a component model, inserting an unknown relation before algebraic elimination can retain a named component’s inputs, outputs and connections. Adding a correction after elimination can be simpler, but the correction then acts on the transformed relations. Recover how it affects the properties needed by the original question. A reduced model may use MMP.9 to derive the contribution its simplification leaves open.

Represent how observations are produced. MMP.7 derives a probability law for records when probability is needed; C.16.IR uses the supplied indication relation to obtain compatible cases or bounds. Fitting an unobserved internal term as if it were measured supplies an extra premise. If that premise is unavailable, fit or constrain through the observable relation instead.

For a dynamic or implicitly defined model, obtain the consequence through its coupled equations and conditions. A good component fit does not settle whether the resulting evolution, initialization or constraints are usable. Apply the mathematical and computational method appropriate to the stated consequence; C.29.2 helps formulate its obtaining operation.

MMP.11:4.5 - Determine which remaining differences matter

Ask what the available observations constrain: the adjustable parameters, the unknown function over a stated domain, or a particular consequence. Use C.16.IR on the resulting observation relation. The function can remain undetermined away from the observed inputs even when its recorded values are fixed.

When ambiguity could change the answer, construct two admitted candidates with the same relevant observations and different receiving consequences. Such candidates show what further information must distinguish. If all compatible candidates or a sufficient bound give the same answer to the present question, use that answer without resolving unrelated differences.

Repeated numerical fits can discover alternatives. Agreement of finitely many fitted instances leaves unsearched alternatives possible. A claim of uniqueness needs its mathematical or statistical grounds; a sufficient decision can require much less. Numerical search failure also differs from a proof that the family is inconsistent with the observations.

Change the question explicitly when a new use requires it. An intervention may distinguish models with the same observational behavior. C.28.MR supplies the replacement of the affected mechanism under its causal premises. Explanation or modification of a working method can require structure beyond that needed for prediction; characterize the required explanatory use through C.2.8 and Explanation Design (EXD).

MMP.11:4.6 - Use the consequence or revise the family

Return the result with the family, input range and conditions that affect its use. It may be a candidate relation, a bound, a conditional prediction, a supported instruction or a located missing contribution. When the model is used to change a working method, C.32.MWA.PP receives the consequence and the conditions the proposed arrangement must preserve; MMP.22 supplies construction when the transformed mathematics does not yet yield a working rule.

Revise the part whose restriction prevents the needed result: the subject premise, permitted dependence, representation, observation relation or obtaining method. Use B.5.RR to carry a changed premise or question through the reasoning. General comparison, portfolios and improvement use the existing C.16, C.11, E.22 and E.23 methods when those questions arise.

Stop when the receiving use has a sufficient answer or the missing contribution is clear enough to obtain. Use C.11.DUA when deciding whether another observation, a richer family or further computation can improve that use enough to warrant its cost.

MMP.11:5 - Archetypal Grounding

MMP.11:5.1 - Construct a response from its known shape

A normalized input u lies in [0,1]. The subject account supports a nondecreasing response r with r(0)=0 and r(1)=1. Its intermediate shape is unknown. Begin with these properties, rather than choosing a straight line as the only candidate.

Choose an integrable h with h(u)>=0 almost everywhere and H=integral_0^1 h(v) dv>0. Define

r(u)=integral_0^u h(v) dv / H.

The endpoints follow by substitution. For u2>=u1, the difference is the nonnegative integral of h over [u1,u2], divided by H. Thus every member is nondecreasing. These curves are absolutely continuous. Every absolutely continuous nondecreasing response with these endpoints has such a representation using its almost-everywhere derivative, but a jump response is outside this family. The needed regularity must come from the question or remain a declared restriction.

For a small calculable family, use linear segments through (0,0), (1/4,q), (1/2,1/2) and (1,1). Their slopes are 4*q, 2-4*q and 1. They are nonnegative exactly when 0<=q<=1/2. Varying q over this interval gives admissible candidates even before any observations are supplied.

Suppose observations establish only the three values at 0, 1/2 and 1. Every q in that interval agrees with them. The consequence r(1/4)<=0.6 follows for this whole family; it also follows for every nondecreasing response with the given midpoint. There is no need to identify q for that question.

Now the receiving use asks whether r(1/4)>0.3. Candidates q=0.2 and q=0.4 satisfy the same observations and give opposite answers. Another repetition at the three old input values does not distinguish these ideal candidates. An observation near the disputed input may help; its precision and cost belong to that new question. Alternatively, a supported additional shape relation could narrow the family.

MMP.11:5.2 - Retain an exchange balance without inventing its mechanism

Two nonnegative amounts x and y exchange a conserved total N. The forward and reverse rates are unknown. Use locally Lipschitz nonnegative rate functions a(x,y) and b(x,y), defined on a neighborhood of the nonnegative states being used, and construct

q=x*a(x,y)-y*b(x,y),

x_dot=-q, y_dot=q.

Adding the two equations gives zero change in x+y for every admitted a and b. At x=0, x_dot=y*b(0,y)>=0; at y=0, y_dot=x*a(x,0)>=0. With these regularity conditions the continuous-time solution preserves nonnegativity. These are properties of the coupled construction. The applicability of conserved exchange to the subject remains a premise.

Even complete knowledge of q need not identify the gross transfers. For any nonnegative locally Lipschitz h, define

a_new=a+y*h, b_new=b+x*h.

The two added contributions to q are x*y*h and -y*x*h, which cancel. The whole observed evolution is unchanged. Every admissible h preserves q.

For a dimensionless instance, take a=b=1. The alternative h=1 gives a_new=1+y and b_new=1+x, yet both models have q=x-y. At x=2, y=1 they both predict x_dot=-1.

Change the question: a proposed intervention suppresses only the reverse transfer while leaving the forward rate law applicable. Under that causal premise, C.28.MR replaces the reverse contribution by zero. The first model gives x_dot=-2; the second gives x_dot=-4 at the same state. Ordinary observations of x and y under the unchanged mechanisms cannot choose between these accounts. A prediction under the old operation can still be useful; the proposed intervention needs a contribution that distinguishes the mechanisms or a sufficient bound covering them.

For dimensional quantities, a and b have inverse-time units, while h has inverse-amount-inverse-time units. Restoring units prevents treating the added terms as arbitrary dimensionless corrections.

MMP.11:5.3 - Construct probabilities while keeping boundary outcomes

A report has three possible outcomes. Let w_i>=0 and let their sum W be positive. Set p_i=w_i/W. Every candidate has nonnegative probabilities summing to one. Conversely, every probability vector on these outcomes is represented by choosing w=p. Thus this construction includes zero-probability outcomes.

The weights (0,1,3) and (0,2,6) both give probabilities (0,1/4,3/4). The parameter vector is redundant even if the probability vector becomes fully determined. There is no need to distinguish those weights when the receiving question uses only the law.

A strictly positive parameterization, such as exponentiating every finite unconstrained parameter before normalization, excludes zero probabilities. It can approximate a zero closely but cannot express it with finite parameters. If the subject account rules out the first outcome, retain that zero in the construction and normalize weights for the remaining outcomes. Whether a very small nonzero value would suffice depends on the receiving question.

The constructed p is a family member, not yet an estimate from data. MMP.7 composes it with selection, rounding or other recording behavior. The appropriate statistical method then determines what the observations support. Changing the recording procedure can change that inference without changing the underlying outcome family.

MMP.11:6 - Bias-Annotation

Familiar formulas can turn an assumed shape into an unnoticed restriction. Flexible fitting can conceal a different commitment: the selected inputs, architecture and loss still determine which functions can be obtained. Recover those choices when they affect the receiving result.

A subject law can also be applied outside its range or boundary. Preserve its grounds and allowed discrepancy. If no supported structural restriction is available, an unrestricted family can be a reasonable candidate for a bounded use; further structure must earn its place through the subject question.

MMP.11:7 - Conformance Checklist

  • The unknown contribution has interpretable arguments, result, domain and permitted dependence.
  • The retained relations have subject grounds and an application range.
  • The construction shows why admitted adjustable values preserve the required properties.
  • Restrictions introduced by representation are carried into conclusions that depend on family coverage.
  • Coupled effects and the relation between the modeled quantities and the observations determine the fitting or inference question.
  • Remaining parameter, function and consequence ambiguities are distinguished when they change use.
  • A changed intervention or receiving question reopens the relevant contribution.
  • The result can be used, qualified or passed to a named next method without requiring unrelated identification work.

MMP.11:8 - Common Anti-Patterns and How to Avoid Them

FailureConsequenceRepair
Fit each contribution independently despite a shared identityThe fitted whole can violate the identity.Build the shared quantity or joint condition into the family.
Treat a penalty as an enforced relationA small fitting loss can conceal a consequential violation.Match the construction and obtaining method to the allowed discrepancy.
Treat a finite fitted family as all admissible relationsA failure inside it becomes an unsupported impossibility claim.State its restrictions and widen or bound the family when the question needs it.
Identify parameters when only a consequence is neededWork is spent resolving distinctions that do not change use.Apply the observation relation to the receiving consequence.
Transfer an observational fit to an intervention without modeling the mechanism changeModels agreeing on observed behavior can imply different intervention effects.Construct the mechanism replacement and the alternatives it can distinguish.

MMP.11:9 - Consequences

The family carries usable knowledge through variation and fitting. It can supply a bound before a particular model is selected, or reveal why more observations of the same kind will leave the important ambiguity intact.

The cost is constructing and checking the representation. Strong restrictions reduce the search but can exclude useful candidates. A flexible family can retain more possibilities while increasing inference cost and leaving more uncertainty. Compare these costs against the result the work needs.

MMP.11:10 - Architectural Rationale

Constructing the free part through the retained relations makes the reason for a property inspectable. A shared transfer preserves a total because the same quantity enters with opposite signs. A shape-constrained response preserves monotonicity because its increments are integrals of nonnegative values. Those reasons remain available when coefficients or learned functions change.

Separating function, representation and receiving consequence also permits economical inference. Many representations of the same function need not be distinguished. Functions that agree on the needed consequence may remain as alternatives. A new intervention can make a formerly irrelevant difference decisive.

The insertion point is therefore an architectural choice in the model. It determines what the adjustable contribution can change, which relations constrain it and which results still have the interpretation the work needs. A symbolic expression and a trained network can each participate in this construction when their mathematical role is recoverable.

MMP.11:11 - SoTA-Echoing

How much of the relation should be left free? The historical SINDy work, Brunton, Proctor and Kutz (2016), Discussion and Appendix B, binds sparse discovery to the chosen coordinates and function library. The universal differential equations construction, section 2.3, combines retained mechanisms with an adjustable function. Sections :4.1-4.3 adopt this choice of where freedom belongs. A small fixed family can be sufficient when its restrictions fit the question. The flexible construction trades additional representation and inference work for retaining variations the small family omits.

How should a required relation survive fitting? For the conservation question in :5.2, :4.2 selects a shared transfer with opposite signs over independently fitted change laws for x and y with only a finite conservation penalty. The latter can fit observations while violating the required total elsewhere. The shared construction preserves that total for every admitted choice of its rate functions. Accept the extra derivation and restriction to conserved exchange in return for that identity; obtain the subject grounds for conservation first. When the question allows a specified discrepancy, a penalty or simpler approximate relation can be sufficient at lower construction cost. Carry the allowed discrepancy into the requested consequence. Direct constraints enforcing the relation remain another option under MMP.10.

Where does the adjustable part enter? Dyad’s model-discovery documentation supports component-level insertion before structural simplification. Micluta-Campeanu and colleagues (2026), sections 2.1-2.2, demonstrate post-simplification correction followed by optional reduction and symbolic replacement. Section :4.4 retains both placements, chosen by their effect on the needed relations. Their thermal application illustrates the post-simplification approach.

What does fitting resolve? Loman and Baker (2025), sections 3.1, 3.3, 3.5 and B.5, distinguish functions, parameters and predictions. Section 3.1 also constructs an algebraic compensation between an unknown function and a mechanistic parameter that preserves observed dynamics. Adopt that construction of indistinguishable alternatives in :4.5; :5.2 adapts it to coupled directional rates. Their finite fitted-ensemble comparisons in :2.4 and Appendix B can reveal alternatives, but agreement of sampled fits does not prove uniqueness over the admitted family. Constructing two admissible alternatives with different receiving consequences establishes the consequential ambiguity, without fitting an ensemble.

Revisit the chosen family when new subject knowledge changes its restrictions, a different observation changes what is distinguishable, a new use needs a formerly discarded difference, or another construction supplies the needed result at lower cost.

MMP.11:12 - Relations

  • MMP.10 constructs representations and the conditions making them admissible. MMP.9 derives an unknown contribution caused by reduction and can use a constructed family to replace it.
  • A.3.3.TR composes a rule of change; C.29.1 relates model consequences to their receiving use; C.29.2 constructs the needed computation.
  • MMP.7 supplies the recording probability law. C.16.IR determines what the resulting indication relation resolves. MMP.8 uses the available information in a choice question.
  • C.28.MR constructs the changed mechanism for an intervention. B.5.RR revises reasoning after a changed premise or question.
  • C.2.8 and Explanation Design (EXD) characterize the explanatory contribution needed by a reader. MMP.22 constructs the working rule from a mathematical change, while C.32.MWA.PP and A.3.1.MV compare its practical arrangement and retain its consequential variation.
  • C.11.DUA compares a further modeling contribution with its cost; general model comparison and improvement use the existing framework methods.

MMP.11:End

MMP.8 - Formulate Information-Dependent Choices Mathematically

Type: Method pattern Status: Stable Normativity: Normative within the stated use

MMP.8:1 - Problem frame

Use this pattern when a mathematical problem mixes quantities you may choose with quantities you do not control, or when a proposed solution depends on information that arrives too late. A solver can find a configuration for each possible circumstance while leaving you unable to select one in advance. A simulation can show a successful continuation because it chose an environmental value that the acting system cannot choose.

Begin with one proposed action. Ask what will be known when it must be chosen, which quantities remain outside that choice, and what result the action must achieve. Construct two possible circumstances that look the same at that moment. If the proposal assigns different actions to them, it needs another observation, a different decision time or a different policy.

The result is a mathematical formulation of the available choices and the requirement they must satisfy, including their dependence on information. It can establish a feasible fixed choice, a policy, a counterexample to the proposal or the missing contribution. This is a general modeling method for design, prediction and control questions. Use it to formulate the question before selecting a solver or a control algorithm.

The finite examples need elementary sets, inequalities and the meanings of ‘there exists’ and ‘for every’. More demanding cases can require optimization, stochastic processes or control theory. If the answer condition and available information are already correctly formulated, use C.29.2 or the applicable computational method to obtain the result.

MMP.8:2 - Problem

A mathematical unknown can represent a decision, an unobserved state, an external input or a quantity constrained by other relations. Treating every unknown as a free decision lets the solution change the problem’s circumstances to make the requested result possible.

Timing introduces another error. A family of solutions indexed by the true circumstance can be mathematically valid while requiring an observation that the acting system never receives. Separately optimizing every future case then grants foresight that the described method lacks.

Which quantities can be chosen, on what information, and what result must the choice achieve? The quantifiers in the formulation must express those dependencies.

MMP.8:3 - Forces

ChoiceConsequence
One decision or an adaptive ruleAn adaptive rule can use later observations but needs a realizable way to receive and act on them.
Possible success or required successA successful case can support possibility while leaving a guarantee unresolved.
Unknown state and random stateAn uncertainty set permits several values; probabilities require an additional model.
Strong requirement and useful feasibilityA guarantee across an oversized circumstance set can reject useful choices; changing the set changes the claim.
Mathematical existence and an obtainable ruleA policy’s existence can matter before an affordable construction is known.

MMP.8:4 - Solution

Separate choices from circumstances, express the relations that must hold, and state which information each choice may use. Formulate the requested result over those objects, then test the formulation against a small contrasting pair of circumstances. Preserve any useful conditional answer when a stronger requirement is unavailable.

MMP.8:4.1 - Name choices, circumstances and consequences

Start from the question in the work. Identify what a participant can change and what is supplied by the subject or environment. Let a denote the choice and w the circumstance. Give both their domains and meanings. A choice can be an arrangement, an input, a rule for later actions or another object the work can construct.

Use the subject relations to connect them to a result. For a deterministic model, a consequence may be written y=f(a,w). An implicit relation R(a,w,y) can retain several possible consequences. Additional unknowns can describe forces, flows, internal states or other quantities jointly constrained by the model. Acausal equations can constrain these quantities without making them free controls.

State what the question requires of the result. If it asks whether every permitted behavior meets a condition, a solver’s ability to find one favorable y does not settle it. If the work can select among the permitted consequences, represent the mechanism that gives it that choice. A.3.3.TR supplies the relevant state-change or interaction rule.

MMP.8:4.2 - Recover the order and availability of information

Describe what is observed before each decision and what arrives afterward. Express the available observation as h(w) when it is a deterministic description of the circumstance. It may reveal only part of w. For a random or noisy observation, specify its probability law for each admitted w. An unknown fixed w can index a family P_w of observation laws without having a probability distribution itself. A joint law is needed when the question also treats w as random and averages over it. MMP.7 constructs the recording law.

A policy pi selects an action from the information available: a=pi(h(w)). Two circumstances with the same h(w) must therefore receive the same action. This uses MATH.2’s condition that an answer remains constant on cases identified by a description. Here the observation determines those groups. The information restriction is often called nonanticipativity: the policy does not use distinctions the decision-maker has yet to observe.

For repeated interaction, use the observation history available at each decision. A policy may remember earlier observations or actions. Omitting that memory is a substantive model choice. Global termination is unnecessary when the question concerns a continuing response; specify the response or progress condition that matters under the admitted inputs.

If an earlier action changes what can be observed, make h depend on that action and describe the cost and timing of observation. If action also changes the circumstance distribution or evolution, include that relation. C.28 supplies the causal-use question and the policy’s permitted pre-action information; C.28.MR constructs the mechanism replacement. A changed distribution cannot be inferred from a favorable selection of historical cases alone.

MMP.8:4.3 - State the required quantifiers and performance criterion

For a deterministic success condition G(a,w), these are different questions:

Working questionMathematical statement
Is there some successful combination?There exist a and w with G(a,w).
Does every circumstance have a successful action, if it were known?For every w, there exists a with G(a,w).
Can one action be selected now that succeeds throughout the admitted circumstances?There exists a such that, for every w, G(a,w).
Can an observation-dependent rule succeed throughout those circumstances?There exists an allowed policy pi such that, for every w, G(pi(h(w)),w).

Choose the statement from the work’s requirement. The first two can expose possibilities or limits even when the latter two fail. A stronger claim can require a different action or information source.

When the receiving use permits failures with a stated probability, name the event and the randomness over which that probability is calculated. For unknown fixed w and random observation O, the question may require P_w(failure of pi(O)) <= epsilon for every admitted w. If w itself is modeled as random and the question concerns performance averaged over circumstances, use the joint law of w and O. An expected cost needs the same choice of what is averaged. Listing possible circumstances specifies neither their probabilities nor the objective. Feasibility, expected performance, tail risk and worst-case performance are alternative questions with different consequences. Use the common choice and characterization methods to decide which matters to the work.

Preserve dependence within the circumstance set. Independently combining several ranges can create impossible circumstances and overstate a requirement. Conversely, excluding an inconvenient circumstance changes the range of the conclusion and needs a subject reason or an agreed narrower use.

MMP.8:4.4 - Construct or refute a usable choice

For each circumstance w, let A(w) be the actions satisfying the requirement under the modeled relations. For a fixed robust choice, seek an action in the intersection of A(w) over the admitted circumstances. An empty intersection refutes that fixed-choice requirement.

For an observation-dependent choice, group circumstances by the observation they produce. For each obtainable observation o, intersect A(w) over the circumstances with h(w)=o. An action in this intersection works throughout that observationally indistinguishable group. In a finite problem, choosing one such action for each observation constructs a policy. One empty intersection proves that this observation cannot support the required policy.

For a noisy observation and an all-cases guarantee, construct the allowed pairs (w,o) from the observation mechanism’s admitted realizations. This relation, rather than a positive probability for each individual report, determines compatibility. For example, if O=w+E with E uniform on [-1,1] and all errors in that closed interval are admitted, the boundary error E=1 remains in an all-cases guarantee despite having probability zero. An almost-sure or specified-probability requirement is a different claim; state that choice.

For a received report o, first check that at least one circumstance is compatible with it. If none is compatible, return the conflict between the report and the observation model under C.16.IR:4.4. A universal statement over an empty set supplies no guarantee for the situation that produced the report. With a nonempty compatible set, intersect A(w) over its members. For a probability-of-success requirement, instead calculate the event under the conditional family or joint law chosen in :4.3; the set of allowed pairs alone supplies no probability weights.

For infinite spaces or long-running interaction, these relations still specify the question, but a usable policy requires the corresponding mathematical and computational construction. A pointwise existence argument does not automatically provide an effective rule. C.29.2 addresses that obtaining work; the present method retains the information restrictions in what it asks the computation to produce.

When several outcomes remain possible for one action and circumstance, apply the required quantifier to those outcomes as well. Derive or inspect a violating outcome when refuting a guarantee. If the missing factor is an unmodeled selection mechanism, obtain it or retain the conditional answer rather than allowing the solver to invent a favorable mechanism.

MMP.8:4.5 - Interpret the result and change the working method

Follow the resulting action or policy through one modeled case and a consequential change. Return the consequence to the original requirement. If a planned response depends on a distinction absent from the available observation, revise the observation, defer the decision, choose a more tolerant action or change the stated goal with the responsible party.

Compare these options by what they change and cost. A more informative observation can enlarge the feasible policy set, but it may arrive too late or cost more than a sufficient fixed choice. C.11.DUA supplies that comparison. An infeasible guarantee can still leave a useful bounded, conditional or risk-qualified proposal.

For a working-method change, explain who or what supplies the observation, what result it provides, when it becomes available and what the receiving action does with it. B.1.5.CM constructs the proposed Method whole; C.32.MWA.PP connects its execution with the available performers and provision. These connections show how the participants can obtain the observation and act on it.

MMP.8:5 - Archetypal Grounding

MMP.8:5.1 - Choose a preload before or after learning an external load

An ideal static arrangement has a downward load w, an adjustable upward preload a and a residual y=w-a. A later use requires abs(y)<=1/4. The load is either 1 or 2, and a can be any value from 0 to 2. The relation is a supplied illustrative mechanical model; use it here to formulate the choice and the information available before it.

If w=1, acceptable choices form [3/4,5/4]; if w=2, they form [7/4,2] after the actuator limit is applied. Both sets are nonempty, so each known load has a feasible choice. Their intersection is empty. No one preload chosen before distinguishing the loads can meet the requirement for both.

Suppose a reading available before adjustment reports which of the two loads is present. The policy a=w then meets the requirement. A reading received only after the preload is locked cannot support that policy at the relevant decision.

Now the tolerance is relaxed to 3/5. The acceptable intervals overlap from 7/5 to 8/5, so a=3/2 works before any reading. Additional measurement is unnecessary for this revised requirement.

MMP.8:5.2 - Decide which participant gets a scarce resource

Two work requests, L and R, may need the only available resource. Exactly one needs it. Allocation succeeds when the resource goes to that request. The instruction must choose L or R deterministically from one report received before allocation.

With no distinguishing report, there are two constant instructions: always allocate to L or always allocate to R. Each fails in one circumstance. A truthful timely report permits an instruction that follows it and succeeds in both. A report arriving after allocation cannot supply that choice.

Now the timely report is noisy. The observing procedure independently chooses one of two channels with equal probability, then sends its L/R report without naming the channel. The model supplies these conditional reporting probabilities; the remaining probability in each row produces the opposite report:

ChannelActual request needing the resourceProbability of a correct report
1L0.9
1R0.5
2L0.7
2R0.9

Use MMP.7 to remove the unrecorded channel by summing over it. For fixed circumstance L, P(report L)=0.5*0.9+0.5*0.7=0.8. For fixed R, P(report R)=0.5*0.5+0.5*0.9=0.7. These conditional probabilities require no assumption about how often L or R needs the resource.

There are four deterministic instructions from one two-valued report:

InstructionSuccess probability in fixed LSuccess probability in fixed R
Always allocate to L10
Always allocate to R01
Follow the report0.80.7
Choose opposite to the report0.20.3

If the requirement is success probability at least 0.65 in each admitted circumstance, following the report satisfies it. None of these instructions gives a zero-failure guarantee. The conditional laws are sufficient to make both statements while the circumstance remains unknown and fixed.

Change the question to average success in a stream of requests modeled as L with probability 0.95 and R with probability 0.05, retaining the channel procedure. Following the report gives 0.95*0.8+0.05*0.7=0.795. Always allocating to L gives 0.95; always allocating to R gives 0.05, and choosing opposite to the report gives 0.205. The instruction with greatest average success among the four is now always L. It still fails in fixed R.

Choose the performance requirement from the work’s purpose before adopting an instruction. The observing model supplies conditional probabilities; the decision about the work determines whether performance in each circumstance or an average matters.

To retain the zero-failure requirement, the work could obtain a truthful report in time or provide enough resource to serve both requests. Compare the cost of those changes with the consequences of accepting a mistaken allocation, using C.11.DUA. A changed channel, recorded channel identity or permission to randomize the instruction changes the information or choice set; formulate the revised question accordingly.

MMP.8:5.3 - Find a strategy, rather than an answer chosen with future knowledge

A program repeatedly receives a bit b and emits a bit a. A requirement asks it to emit the same bit. If receipt precedes emission, the rule a=b works on every round. If emission must precede receipt, each possible future bit has a matching answer, but no deterministic rule using only the earlier history can guarantee a match against every admitted next bit.

To see the failure, hold the earlier history fixed. The rule chooses either 0 or 1. Both next input bits are still admitted, including the opposite one. That continuation refutes the guarantee for this history. Inspecting more successful traces cannot remove it.

A changed requirement may ask for success probability under independent fair input bits. Any earlier choice then matches with probability 1/2 in one round, including a randomized earlier choice independent of the next bit. Success in every one of N such rounds has probability 2^(-N). These probabilistic claims use the new input assumption. They do not establish success against every input stream.

This is a continuing computational interaction. The same timing constraint applies to a team or controller: a response can use only information available before it acts.

MMP.8:6 - Bias-Annotation

Optimization tools encourage viewing every variable they can assign as an available choice. Recover the subject meaning of each unknown before interpreting a solution. A second bias is to make robustness the default goal. State the receiving requirement first; a conditional answer or an explicitly accepted risk may be more useful than an infeasible all-circumstances guarantee.

MMP.8:7 - Conformance Checklist

  • Are choices, circumstances and consequences distinguished by what the acting system can change?
  • Does each choice depend only on information available at its decision time?
  • Do the quantifiers express the intended possibility, guarantee, policy or probabilistic question?
  • Are dependent circumstances and remaining possible outcomes preserved in the formulation?
  • Does a witness, empty intersection, policy or bound have the claimed meaning?
  • Does the interpreted result support an action, a changed method or a specific unresolved contribution?

MMP.8:8 - Common Anti-Patterns and How to Avoid Them

Failure in this workRepair
A solver changes an external load or unknown state to satisfy the goal.Treat it as a circumstance and apply the required quantifier.
A separately optimal action for every future is presented as one available strategy.Group cases by the information available at the decision; require a common action within each group.
An existentially chosen consequence stands in for all allowed behavior.Retain the behavior relation and test the quantifier required by the goal.
A larger uncertainty set is called safer without examining its subject meaning or cost.Check the joint possible circumstances and choose the requirement for the actual use.
Better observation is required after a sufficient common action is already available.Compare what the additional information can change before commissioning it.

MMP.8:9 - Consequences

The formulation prevents computation from supplying unavailable control or foresight. It can expose a useful change to observation, timing or the requirement before optimization begins. It can also be harder to solve than its scenario-wise surrogate; that extra difficulty reflects the work’s requirement.

MMP.8:10 - Architectural Rationale

Quantifier order and permitted dependence express different aspects of the working problem. ‘For each circumstance there is an action’ concerns a family of possible solutions. An executable policy also needs a way to select its action from obtainable information. Grouping indistinguishable circumstances makes that additional condition visible and gives a constructive test for finite cases.

This mathematical formulation can describe physical adjustment, resource allocation or computational interaction. Its source premises and available actions differ across those practices. The method keeps those differences explicit while reusing the same reasoning about choices and information. It complements FPF’s continuation and computation methods by constructing the mathematical answer condition they consume.

MMP.8:11 - SoTA-Echoing

Boyd and Vandenberghe, Convex Optimization, section 4.1, distinguish the feasible set from the objective and transformations of a problem. Adopt that separation in :4.1 and :4.3. The present finite constructions do not require convexity; the book’s convex solution guarantees apply only under their conditions.

Ben-Tal, Goryashko, Guslitzer and Nemirovski, Adjustable robust solutions of uncertain linear programs (2004), develop adjustable decisions alongside choices fixed before uncertainty is revealed. Use this distinction when formulating information-dependent choices. Their tractability results concern restricted adjustable decisions under the paper’s assumptions. They do not extend to unrestricted policies in general.

Duchi, Optimization with uncertain data (2018), sections 1 and 6, compares uncertainty-set requirements with probabilistic ones and makes choosing the uncertainty set a modeling question. Adopt that choice explicitly; no worst-case objective is imposed by this pattern.

Vayanos, Georghiou and Yu, Robust Optimization with Decision-Dependent Information Discovery, version 3 (2022), treats actions that affect when uncertainty can be observed. Carry that extension into :4.2 and the return to method design. Its algorithms address the robust optimization problems studied there.

The interval, allocation and bit-response examples are constructed demonstrations of the method under their stated assumptions.

MMP.8:12 - Relations

B.5.FM and C.29 connect the work’s question to its mathematical formulation. A.22.CGUS exposes allowed continuations; A.3.3.TR constructs the behavior relation. MMP.7 supplies a missing observation law for probabilistic uses. MATH.2 explains the identification of cases that preserve a requested answer; here grouping by available information determines which actions a policy can distinguish. C.29.2 obtains a computation for the formulated question and C.29.3 examines realization. C.28 governs causal claims when acting changes the represented world. B.1.5.CM uses the result to construct the composition of observing, deciding and acting Methods; C.32.MWA.PP compares the needed contributions, timing and provision.

MMP.8.SD refines this formulation for continuing decisions. It constructs a decision-sufficient state or belief, transitions and observations, and the relation between a present choice and its later consequences. Use that refinement when the one-step formulation leaves a consequential continuation unresolved.

MMP.8:End

MMP.8.SD - Construct a Sequential Decision Model from Information and Consequences

Type: Method pattern Status: Stable Normativity: Normative within the stated use.

MMP.8.SD:1 - Problem frame

Use this when a choice changes what can be done or learned later, and comparing its immediate result does not settle the useful course of action. You might reserve a resource for a later request, inspect an unfamiliar document before choosing how to process it, or pursue a target whose attainment depends on several moves.

Start with two possible first actions. For each, write what becomes available before the next choice and what the next choice can change. This small decision tree often reveals the missing resource, remembered event or uncertainty that a proposed “current state” has discarded.

The object being constructed is a sequential decision model: a mathematical account connecting available information, allowed actions, subsequent observations and accumulated consequences. Its result lets a reader compare a present action together with an admissible continuation. A continuation is the later policy: a rule for choosing actions from the information available then.

This is the sequential refinement of MMP.8 - Formulate Information-Dependent Choices Mathematically. It retains that pattern’s separation of chosen quantities, uncontrolled circumstances and information available at each choice. It adds the construction of a state sufficient for the selected decision problem and the relation between current action and remaining consequence. A short history tree is a valid starting model; a compact state becomes useful when the tree repeats questions or grows too large.

A reader can work the finite cases with conditional probability, finite sums and maxima, or with sets of possible outcomes. For continuous, constrained or indefinitely continuing models, the argument requires preparation in stochastic control, decision theory or the applicable field. If you obtain it from another contributor, supply the timing, allowable actions and consequence criterion, and ask for the model and its limits.

Not this pattern when the useful comparison is already a single choice with no consequential continuation. Use MMP.8 directly. If the decision model is adequate and only its solution is expensive, the next work is computational. If the difficulty is deciding whose interests or which consequences should count, use the applicable decision or strategy practice to obtain that choice from whoever has authority to make it.

MMP.8.SD:2 - Problem

A record of the present can omit information that changes a later action’s availability or value. A forecast can predict the next reading correctly while omitting an accumulated loss that matters at the end. A plan can also give its future decision maker information that will arrive too late.

These defects can survive accurate computation. Optimizing a model of the latest observation, a guessed hidden state or immediate payoff may return the best answer to a different problem. Conversely, retaining every observation and solving for every possible future can cost more than the decision warrants.

The modeling question is: what information and consequence account make the present choice and its continuation comparable, under the conditions of use?

MMP.8.SD:3 - Forces

ForceTension
Useful memoryA compact state saves work, but merging histories can remove an allowed action or change its consequences.
Information timingLater observations support conditional action, while a plan chosen now cannot depend on their unrealized values.
Present and later consequencesA locally attractive move may consume a resource, remove an option or change what can be learned.
Criterion choiceExpected total gain, probability of meeting a target and worst possible loss can prefer different policies.
Affordable adequacyExact reconstruction of the hidden state may be unnecessary; an adequate comparison or bound can settle which action to take.
Model and computationA correct solver can answer a deficient model, and an adequate model can remain computationally difficult.

MMP.8.SD:4 - Solution

Construct the information available at each choice, retain what determines the relevant continuation, and derive the accumulated consequence of a policy. Use that derivation to locate the premise responsible when a changed condition alters the answer.

MMP.8.SD:4.1 - Fix the decision times and the consequence being compared

Recover from MMP.8 the choices, uncontrolled circumstances and information available before each choice. Mark the order of observation, action, transition and consequence. Include delays, commitments and stopping opportunities when they change what can be done.

For decisions at times t=1,…,T, let H_t be the available history immediately before action A_t. It contains received observations and known previous actions. Include a realized gain or cost in this history only if the decision maker can know it then. A policy π_t(H_t) selects an allowed action from that history; randomization is an additional allowed operation when the problem permits it.

State the consequence criterion supplied by the receiving use. One common finite model maximizes

J(π) = E^π[R_1 + R_2 + … + R_T + G].

Here R_t is the gain at decision step t, G is a terminal contribution, and the expectation uses the trajectory law induced by policy π and the model. Gains can be negative costs. Their addition presupposes a common meaning and scale. Specify the initial information or distribution under which policies are compared.

An expected sum is one choice of criterion. For a target such as “finish with at least k points,” the consequence can instead be the terminal indicator, equal to 1 when the target is met and 0 otherwise. Its expectation is the attainment probability. For several criteria, retain their comparison or trade-offs until an applicable choice method supplies a selection rule. A scalar “reward” does not decide that rule.

Use a horizon or stopping condition appropriate to the work. Assign the consequences left at the horizon, such as unused stock or unfinished obligations. A discounted sum gives later gains smaller weights; choose that meaning deliberately. Shortening a computational planning window does not make omitted consequences disappear.

MMP.8.SD:4.2 - Construct how an action changes the world and the available information

Write the transition and observation account for each allowed action. For a finite stochastic model, choose an underlying state X that retains the information needed for the transition account. One possible representation is the joint law

K_t(x', y, r | x, a)
  = modeled probability of next state x',
    next observation y and current gain r
    when action a is applied in state x.

This representation assumes that, given x and the applied action a, the omitted history does not further change the law. Retain a missing historical distinction when that assumption fails. Using a joint law permits dependence between the transition, observation and gain. Factor it into simpler laws only when the model supports the corresponding conditional independence. A deterministic rule or a relation of possible successors can replace probabilities when that is what the available knowledge warrants.

Identify the source of these relations. A subject model supplies the consequences of applying the action; MMP.7 supplies how observations are recorded and become available. An intervention model through C.28.MR can supply the changed mechanism. A conditional association among logged actions and outcomes is not automatically the law under a new policy.

An action can change both the subject and what the next decision maker will know. An inspection might consume time, disturb an object and produce a reading. Include each effect that changes the policy comparison. If two participants receive different observations, preserve their separate information; a policy using all their private information would require a means of sharing it before the choice.

For a small problem, enumerate the allowed actions and possible next observations at each history. Label branches with probabilities or allowed circumstances and gains. This tree supplies a reference calculation before any compression.

MMP.8.SD:4.3 - Retain a state sufficient for the selected continuation

Propose Z_t=s_t(H_t), a summary obtainable from the information available at time t. It may be an observed state, a history window, a set of possibilities or a belief: a probability distribution over the hidden state conditional on the available history.

When the retained state is a belief, A.3.3.PI:4.4 supplies its prediction and observation update under a specified model. Use that distribution and update here. Include uncertain fixed parameters or other remembered quantities when they affect future consequences. A point estimate can discard differences that change an action’s value. A posterior over physical coordinates alone can also be insufficient when a resource budget or terminal target depends on the past.

For the expected additive criterion in :4.1, a constructive sufficient test compares any two admitted histories at the same decision time having the same proposed Z. For every action covered by the model, determine whether those histories give:

  • the same allowed actions;
  • the same conditional expected current gain;
  • the same conditional law of the next retained state after the action.

At the terminal time, the conditional expected terminal contribution must also depend only on the retained state. Supply an initialization and an update from retained information, the action taken and the next received observation. These conditions let the continuation calculation use Z in place of the full history. A decision may need less information than this sufficient test requires.

The comparison covers the actions and histories for which the policy is to be used, including alternatives to the former policy. A match only along recorded behavior can hide distinctions that another action makes consequential. Predicting irrelevant observations is unnecessary when their differences affect neither the criterion nor future choices.

When the test fails, identify the missing distinction and repair the state. An unspent resource, elapsed time, accumulated amount or belief over an unknown mode can be the needed addition. If the repair is expensive, retain the history tree, restrict the claimed use, or use a bound sufficient for the present comparison. Two histories with different forecasts can still support the same action when that action dominates under both.

For an approximate summary, determine how its errors can change the action comparison. Carry errors in current gains and in expected continuation through the selected horizon, using an applicable bound or model criticism. If computed action values each have justified absolute error at most e relative to the intended model and criterion, a largest value more than 2e above every rival has the same maximizing action. Without such separation, preserve the unresolved comparison or improve only the approximation that can change it. A small prediction error on a training sample alone does not supply this guarantee.

MMP.8.SD:4.4 - Derive current gain plus continuation

First evaluate a policy that selects its actions from the retained state. With a sufficient state, let V_t^π(z) mean expected gain from time t onward when the policy is followed, conditional on current retained state z. For a deterministic policy, the finite model gives

V_(T+1)^π(z) = g(z)
V_t^π(z) = r_t(z, π_t(z))
           + sum_z' P_t(z' | z, π_t(z)) V_(t+1)^π(z').

Here g(z) is the conditional expected terminal contribution, r_t(z,a) the expected current gain, and P_t the next retained state’s law derived from :4.2–:4.3. A randomized rule averages the right-hand side over its action probabilities.

The equation follows by splitting the accumulated gain into the current term and the remaining terms, then conditioning on the next information state. Thus each action is compared with what can follow it, including the information then available.

For a finite state and action problem with nonempty allowed action sets, whose only policy restrictions are those local sets, the best attainable value satisfies

V_(T+1)(z) = g(z)
Q_t(z,a) = r_t(z,a) + sum_z' P_t(z' | z,a) V_(t+1)(z')
V_t(z) = max over a in A_t(z) of Q_t(z,a).

An attaining action at each reached state defines an optimal policy for this model and criterion. Finite sets make these maxima attainable. For more general spaces, determine the relevant existence, measurability and integrability conditions. When a maximum is not attained, distinguish the supremum from any obtained approximate policy.

The order of choice and averaging matters. When a later observation is available before the next action, its branch can use its own continuation. A fixed action sequence cannot use that observation. At a common decision node, selecting a different action for each still hidden state would add unavailable information.

Retain restrictions coupling choices across histories or times. A total resource limit can often be represented by remaining resource in Z; a constraint on the whole policy may instead require a constrained formulation. Independently maximizing every node can violate a coupling that the state has omitted.

These equations define the mathematical continuation problem. CMP.3 supplies sharing and scheduling of repeated subcomputations; CMP.9 supplies a sampling procedure and the error of estimated expectations; CMP.5 can supply a relaxation, a usable policy and an improvement bound when its recovery conditions hold. C.29.2 supplies the wider computational formulation, including a continuing response rather than a terminating answer.

For indefinitely continuing use, first select a finite total, discounted sum, average rate or other well-defined criterion. For example, bounded per-step gains and a discount factor γ with 0≤γ<1 make the infinite discounted sum finite. A fixed-point or limiting equation then needs conditions appropriate to that criterion. A solver’s convergence and a policy’s consequences remain separate questions.

MMP.8.SD:4.5 - Match the continuation to the uncertainty and objective

When the available information is a set of possible circumstances, evaluate a policy over the permitted complete trajectories. Compare its worst consequence, an interval or another requested result without inventing a probability distribution.

A stage-by-stage worst-case calculation is justified only if the retained description preserves which continuations remain possible. In particular, one unknown parameter fixed throughout a run cannot silently take a different worst value at each stage. Carry that parameter’s compatible set and any information learned about it, or keep the coupled trajectories. :5.4 shows a choice reversed by losing this dependence.

Likewise, a terminal threshold depends on the accumulated amount. Retain that amount if it is observed; otherwise retain the uncertainty about it together with the other relevant state. In :5.3, current expected gain is enough to compare one objective but insufficient for another.

For a compressed or learned state, separate three possible claims: performance of a specified restricted policy; an optimum within that policy class; and an optimum among all policies allowed by the available history. Convergence for one memory representation establishes only the result justified for that representation and the policies considered. Evaluate the returned policy under the original information and consequence account, with the uncertainty or approximation relevant to its intended use.

MMP.8.SD:4.6 - Return the model and propagate a consequential change

Return the state or belief and its update, admissible policy information, action-dependent relations, consequence criterion and continuation relation. Include a computed policy comparison or bound when it is useful, and state the assumptions that make it applicable. A formula alone can be the right input to a computational specialist; a two-branch calculation can already settle an ordinary choice.

Work through an available case from initial information to the result that the receiving activity uses. If a state reduction was needed, include the histories it would otherwise merge. Compare with a serious simpler option: a fixed plan, immediate-gain choice, full history tree or an existing policy with an adequate bound. Added state and computation earn their cost by changing the answer, making it obtainable or preserving a needed qualification.

When an actual premise changes, follow the affected dependency. A delayed observation changes allowable conditioning; changed resource availability changes actions and transitions; a new target changes the consequence account and may change the state. Recalculate the affected continuations. Use MMP.14 when observations reveal systematic mismatch in the proposed model, and the subject method when its action consequences need repair.

Use C.11.DUA to decide whether additional observation, modeling or computation can improve the receiving use enough to warrant its cost. An existing bound can settle the question. Testing a hypothetical change can expose an assumption that affects the decision.

A qualified contributor can construct the model or calculate its consequences. Applying a policy in the subject requires the observations, permitted actions and performing method that the model assumes. Methodological work uses the model to decide, for example, whether to observe before acting, retain a resource or change a method after an informative result; it also supplies the practical conditions under which that sequence can be performed.

MMP.8.SD:5 - Archetypal Grounding

The following are constructed finite models. Their numbers show what follows from the stated assumptions.

MMP.8.SD:5.1 - Reserve a resource for an observed later request

A portable power unit has enough energy to serve one task in either of two slots, with no recharge. Serving the current task in slot 1 yields 4 units on an already selected gain scale. In slot 2 a different task arrives with probability 0.6 and yields 10 if served. Arrival is observed before the slot-2 decision. Unused energy has terminal value zero; the current task cannot be deferred. Serving consumes the whole unit.

Let b be remaining energy, either 0 or 1, and d indicate the later arrival. At slot 2,

V_2(b,d) = 10*b*d.

Use the retained state (time, remaining energy, current request). The slot-1 comparison is

serve now: 4 + 0 = 4
reserve:   0 + 0.6*10 + 0.4*0 = 6.

The resulting policy reserves energy, then serves if the request arrives. It has larger expected gain than immediate service, although it yields zero when no request arrives. If the requirement is a gain of at least 4 in every admitted case, serving now meets it and reservation does not.

Two histories can show the same later request but differ in whether the energy was already used. Merging them would make the slot-2 service appear available after both histories. Retaining b prevents that error.

Suppose the arrival probability changes to 0.3, with the other premises unchanged. Reservation now gives 3 and immediate service gives 4, so the first action changes. If instead the only available probability statement is 0.55≤p≤0.65, reservation gives expected gain between 5.5 and 6.5 and exceeds 4 throughout. Refining p is unnecessary for that comparison.

MMP.8.SD:5.2 - Pay for information only when a later choice can use it

An unfamiliar encoded document uses format A or B, initially with equal probabilities. Two supplied decoders are available: the matching decoder gives a usable output worth 10, and the other gives 0. The model permits one final decoder application. An optional diagnostic costs 1 on the same gain scale and leaves the format unchanged.

The diagnostic returns + with probability 0.8 in format A and 0.2 in format B; the complementary probabilities give −. Its result arrives before decoder selection. The supplied observation model and A.3.3.PI’s conditioning give:

Information before decoder choiceProbability of format ABest decoderExpected final gain
No diagnostic0.5A or B5
Diagnostic +0.8A8
Diagnostic −0.2B8

The decision state can be (stage, p), where p is the current probability of A. At the final stage,

V_2(p) = max(10*p, 10*(1-p)).

Each diagnostic result has probability 0.5 under the initial model. Thus

skip diagnostic:  V_2(0.5) = 5
use diagnostic:  -1 + 0.5*V_2(0.8) + 0.5*V_2(0.2) = 7.

The continuation’s choice changes with the observation. Fixing one decoder in advance gives 5 before the diagnostic cost, or 4 after it. This is why the extra information has value here.

Now the diagnostic result is delayed until after the final decoder choice. The decoder choice can no longer depend on that result. Using the diagnostic then has net expected gain 4, so skipping it gives the better value 5. This repair changes the information timing rather than the arithmetic of conditioning.

A different changed condition also reverses the comparison: with a timely but weaker symmetric diagnostic that is correct with probability 0.55, using its result gives -1+10*0.55=4.5. Skipping still gives 5 and is preferable.

The decoder and diagnostic behavior are premises of this example. A real use requires the applicable processing and observation methods to supply those consequences.

MMP.8.SD:5.3 - Retain accumulated progress when the terminal goal changes

In a two-round game, all awarded points are observed immediately. The first move is either A, giving 2 points with probability 0.5 and 0 otherwise, or B, giving 1 point certainly. In the last round, the player chooses Safe, adding 1 point certainly, or Gamble, adding 3 with probability 0.4 and 0 otherwise. The gamble’s outcome is independent of the first round.

The objective is initially to maximize the probability of finishing with at least 3 points. Let c be points already earned. The terminal contribution is g(c)=1 when c≥3 and 0 otherwise. With zero intermediate contribution, the continuation compares terminal attainment:

Points before last roundSafe: attainment probabilityGamble: attainment probabilityBest continuation
000.4Gamble
100.4Gamble
210.4Safe

Therefore first move A gives 0.5*1+0.5*0.4=0.7, while B gives 0.4. Choose A and condition the last move on earned points. Retain the state (round, c). The round number alone is insufficient: histories with c=0 and c=2 require different continuations.

If the objective were expected final points, Gamble’s expected increment 1.2 exceeds Safe’s 1 at every c. Both first moves have expected increment 1, so both then give 2.2 expected final points. That calculation answers a different question from attainment probability.

Now change the requested terminal target from 3 to 4 points. From c=2, Gamble attains it with probability 0.4 and Safe fails. From c=0, neither last move can attain it. Thus A gives 0.5*0.4+0.5*0=0.2. B followed by Gamble gives 0.4 and becomes preferable. The transition probabilities remain unchanged; the changed terminal criterion alters both the continuation and the first choice.

MMP.8.SD:5.4 - Preserve one unknown condition across stages

A two-stage processing route incurs costs w and 1-w, where an unknown w is fixed for the whole run and belongs to {0,1}. Choosing that route commits to both stages. A supplied alternative has total cost 1.5. The question is to minimize worst total cost, with no probability model.

The two possible cost sequences for the route are (0,1) and (1,0); both total 1. The route therefore beats the alternative. Adding the worst first-stage cost 1 to the worst second-stage cost 1 gives 2, but combines different possible runs.

For the first choice, the common total 1 already suffices. If the remaining cost is later needed and the first-stage cost is observed, that observation identifies w and determines the remaining cost.

Now the operating condition is allowed to change between stages: costs are w_1 and 1-w_2, with all four pairs (w_1,w_2) allowed. The pair (1,0) gives total cost 2. The route’s worst total is now 2, so the fixed-cost alternative 1.5 is preferable. The old conclusion fails because the admitted dependence changed.

MMP.8.SD:6 - Bias-Annotation

Finite stochastic examples make expectations and recursion easy to inspect, but may encourage a probability model or scalar gain where neither is supplied. Keep non-probabilistic possibilities and multiple criteria when that is what the work supports.

A model with one decision maker can also conceal information held separately by different participants. Represent communication and its timing when the proposed continuation depends on shared knowledge. Learned memory is attractive in large problems; its convenience does not establish sufficiency for a changed action set or criterion.

MMP.8.SD:7 - Conformance Checklist

  • The model states the decision order, information available before each action and allowed policy class.
  • Transition and observation relations distinguish applied actions from uncontrolled circumstances and retain consequential dependence.
  • The proposed state has an obtainable initialization and update. Its sufficiency or approximation is tied to the selected actions, criterion and horizon.
  • The accumulated consequence includes the relevant terminal contribution, resource constraint or remembered progress.
  • The continuation relation preserves the order of observation, choice and aggregation over uncertainty.
  • A worked comparison returns a useful policy consequence, bound or unresolved distinction; any real changed premise is propagated to its affected use.
  • Additional data or computation is selected for its possible effect on the decision. A sufficient result can be used directly.
  • The mathematical model, computed policy and subject performance have distinct claims and required contributions.

MMP.8.SD:8 - Common Anti-Patterns and How to Avoid Them

Anti-patternWhat goes wrongRepair
Latest reading called the stateHistories with different resources or beliefs become indistinguishable.Compare the histories’ allowed actions and continuation, then retain the missing distinction.
Future knowledge used nowA policy selects a branch before its observation arrives.Place observations and commitments on the same decision timeline.
Immediate score treated as total valueResource use or information acquisition changes the future comparison.Include the continuation under each current action.
Every objective written as a sum of immediate rewardsA terminal target or policy constraint changes meaning.Preserve the criterion and augment the state or formulation where it needs history.
Independent worst values substituted for one fixed unknownThe calculation creates a trajectory excluded by the model.Preserve coupling across stages or explicitly adopt the enlarged uncertainty set.
Convergence treated as decision sufficiencyA solver can converge for a representation that loses relevant history.Identify the solved problem and evaluate the returned policy under the intended model.

MMP.8.SD:9 - Consequences

The reader obtains a model that explains why a present action is useful together with what can follow it. It can reveal that a resource should be retained, that a reading is worth obtaining only before a commitment, or that a changed target requires remembering different information.

The construction also locates missing contributions: an observation law, an action consequence, a criterion, a retained distinction or a computational procedure. A small bound may close the comparison before any large policy computation.

A sufficient state can greatly reduce repeated reasoning, but constructing it and solving the resulting problem can still be costly. More detailed modeling can worsen the work when it cannot change the receiving decision.

MMP.8.SD:10 - Architectural Rationale

The construction begins with available histories because they expose which information the decision maker has before choosing. A state is then a justified reduction of the decision problem. Starting from a convenient list of current features reverses that order and can hide lost memory.

Prediction contributes the distribution or set needed to describe a continuation. Decision modeling additionally specifies allowable actions and how consequences are accumulated and compared. In :5.3, identical transition knowledge supports different policies after a change of goal. In :5.2, identical diagnostic accuracy has different value after a change of timing.

A full history tree is a serious alternative to state compression. For a small problem it is often the clearest and cheapest model. A fixed plan is adequate when future information cannot improve the selected comparison; the delayed diagnostic shows such a boundary. For larger problems, a justified compact state makes repeated questions reusable by computational methods.

The continuation relation describes the mathematical problem consumed by those methods. It does not prescribe a solver or the worth criterion. This keeps model construction usable in physical operations, investigation, computational work and method design without replacing their subject methods.

MMP.8.SD:11 - SoTA-Echoing

The practice question is how to retain enough information for a useful sequential choice without demanding an unnecessarily complete reconstruction.

Question and comparisonAdopted or adapted contribution and limit
Can a compact summary replace the full history tree?Adopt the information-state line of Subramanian, Sinha, Seraj and Mahajan, Approximate Information State for Approximate Planning and Reinforcement Learning in Partially Observed Systems, JMLR 23 (2022), §§2.2–2.3 and 3.2, especially Theorems 5 and 9. Preservation of expected gain and the next summary’s law supports the reduction in :4.3–:4.4; approximate preservation needs a consequence bound. A full history tree remains preferable when small. This line saves representation and computation only when its conditions are obtainable; prediction fit alone supplies less. Reopen when a newly relevant action, history or criterion changes the equivalence between merged histories.
What if compact memory is useful but not sufficient?Adapt Sinha and Mahajan, Agent-state based policies in POMDPs: Beyond belief-state MDPs, arXiv v1, 24 September 2024, §§II-C–II-D and III. Its comparison of policy classes supports :4.5: evaluate a restricted controller as such, rather than assuming an arbitrary recurrent memory admits the continuation equation in :4.4. Direct policy search is an alternative when sufficient-state construction is too costly. Preserve whether its result is locally optimal or best within a restricted policy class. Reopen when memory, available computation or required policy class changes.
Does convergence of a modern learning algorithm close the state question?Reject that inference, using Sinha, Geist and Mahajan, Convergence of regularized agent-state-based Q-learning in POMDPs, arXiv v2, 2 September 2025, §§II-B–IV and Theorem 1. Under its learning-rate and visitation assumptions, the limit is for a regularized model that depends on the behavior policy’s limiting distribution. The result sharpens :4.5’s separation of numerical convergence from the intended policy comparison. Regularized learning remains a computational option; it brings its objective and representation conditions. Reopen when a proposed solver claims a stronger use than its result supports.
What if probabilities are unavailable and the criterion is worst consequence?Adapt Dave, Venkatesh and Malikopoulos, Approximate Information States for Worst-Case Control and Learning in Uncertain Systems, arXiv v2, 6 April 2024, §§II–III. Conditional ranges and a criterion-specific continuation provide a worst-case comparison when probabilities are unavailable. The paper derives continuations for maximum instantaneous and terminal cost. For accumulated costs, :4.5 and :5.4 preserve the allowed complete trajectories. The trade-off is a worst-case comparison in place of a probabilistic average; reopen when the uncertainty set or its cross-stage dependence changes.

MMP.8.SD:12 - Relations

  • MMP.8 - Formulate Information-Dependent Choices Mathematically supplies choices, uncertainty, information timing and admissible policies. This nested refinement constructs their sequential state and continuation.
  • A.3.3.PI - Retain the Information Needed for Prediction, especially :4.4, supplies prediction and hidden-state belief updates. This method consumes them in action and accumulated-consequence comparisons.
  • MMP.7 supplies the recording law and available observations. MMP.13 can supply inferred model quantities with their uncertainty; neither contribution alone establishes the effect of a new action policy.
  • C.28.MR supplies intervention through a changed mechanism. A subject method supplies or justifies the resulting action and observation relations used here.
  • MMP.14 supplies criticism and repair when comparable predictions fail. A detected failure here identifies which state, information or consequence premise must return to modeling.
  • C.29.2 - Computational Formulation specifies the computational answer or continuing response and the procedure needed to obtain it.
  • CMP.3 - Share and Schedule Repeated Subcomputations organizes repeated evaluations of the continuation. CMP.9 - Construct a Randomized Estimator or Sampling Procedure estimates expectations with a computational error account. CMP.5 - Bound an Optimum or Recover a Feasible Candidate through a Relaxed Problem supplies relaxation, candidate recovery and improvement bounds for the chosen formulation.
  • C.11.DUA helps decide whether further information or computation is worth obtaining. FPF choice and portfolio methods and the applicable development, Strategy or Core Method-development contribution supply criteria, compare the returned alternatives and organize their practical use.

MMP.8.SD:End