Skip to content

Solvable Baseline Decomposition

Solve the nearest tractable version first, then add only those corrections whose size, order, and validity range can be defended.

Gap-Fill Rationale

solvable_baseline_decomposition was selected as queue position 17 in the accepted-prime gap-fill pilot. It targets the canonical accepted primes decomposition and perturbation_theory, both listed in the queue as actual zero-any targets. The pre-draft check found adjacent accepted archetypes for core-first modeling, bounded approximation, perturbation testing, layered validation, and generic decomposition, but no accepted archetype that centers the full baseline-solution plus perturbative-correction structure.

The draft therefore proceeds as a full gap-fill archetype. It should not be collapsed into core_model_first unless that accepted parent is explicitly expanded to include perturbation-term definition, correction-order sequencing, residual error budgeting, and validity-range boundaries.

Essence

Solve the nearest tractable version first. Then treat the difference between that reference case and the real target as a named, bounded perturbation. Add corrections in an explicit order, test whether they improve the result, and stop or switch methods when the correction logic breaks down.

The archetype is useful because it lets a team preserve what is known exactly or reliably while still acknowledging that the real problem is more complex. It keeps simplicity from becoming denial and complexity from becoming unmanageable.

Compression statement

Solvable Baseline Decomposition treats a complex target problem as a controlled departure from a simpler solved or strongly characterized reference case. It defines the baseline solution, isolates the difference between baseline and target as perturbation terms, orders corrections by scale or importance, recomposes the solution, and checks whether the correction sequence remains valid before using the result.

Canonical formula: target problem = solvable baseline + perturbation terms; solution ≈ baseline solution + ordered corrections + residual error boundary

When to Use This Archetype

Use it when a hard target problem is close enough to a solved or trusted reference case that the target can be represented as baseline plus departures. The baseline might be a known equation, benchmark design, equilibrium model, normal operating plan, reference workflow, or canonical case. The departure might be a forcing term, residual, design modification, exception class, scenario load, environmental shift, or secondary effect.

It is especially appropriate when the user needs an interpretable approximation, a staged path from simple to complex, or a disciplined way to decide which refinements matter.

Structural Problem

A complex problem is too hard to solve all at once, but treating it as merely “too complex” wastes the structure that is already known. Conversely, treating the simple case as the answer ignores important departures. The structural problem is the lack of a disciplined relation among:

  • the target problem,
  • the solvable baseline,
  • the residual differences,
  • the correction sequence,
  • the error budget, and
  • the validity boundary.

Without those relations, teams either overgeneralize the baseline, overbuild an opaque full-complexity model, or bolt on corrections without knowing when those corrections have lost validity.

Intervention Logic

The intervention is to choose a reference problem that can be solved or strongly characterized, show how the target reduces to that baseline when departures are removed, isolate the departures as perturbation terms, and add corrections in an ordered way. Each correction should reduce a named residual or improve target fidelity. The recomposed answer must carry its own validity range.

A typical intervention path is:

  1. Define the target problem and decision tolerance.
  2. Select a solvable reference baseline.
  3. State the baseline-target correspondence rule.
  4. Define perturbation terms as specific departures from the baseline.
  5. Normalize scale to judge whether the departures are small or bounded.
  6. Solve or document the baseline solution.
  7. Add corrections in order of magnitude, dependency, or decision value.
  8. Compare residual error with the acceptable budget.
  9. Declare the validity range and fallback triggers.

Key Components

Core components are the target problem frame, solvable reference baseline, baseline correspondence rule, perturbation delta definition, smallness or validity assumption, correction order sequence, residual error budget, validity range boundary, and refinement/recomposition rule.

Optional components include a scale normalization reference, symmetry or conservation register, benchmark case set, convergence monitor, and nonperturbative fallback path.

The most important invariant is that the baseline remains visible after corrections are added. If the corrected solution cannot explain which part came from the baseline and which part came from a perturbation, the archetype has degraded into generic approximation.

Common Mechanisms

Common mechanisms include zeroth-order model selection, delta term isolation, dimensionless small-parameter checks, first-order correction passes, successive-order refinement, residual comparison tests, convergence or asymptotic behavior checks, benchmark backtests, validity boundary scans, and fallback trigger rules.

These are mechanisms, not the archetype itself. The archetype is the cross-domain structure that links a solvable reference case, controlled departures, ordered corrections, and a declared validity boundary.

  • Benchmark Backtest
  • Convergence or Asymptotic Behavior Check
  • Delta Term Isolation
  • Dimensionless Small-Parameter Check
  • Fallback Trigger Rule
  • First-Order Correction Pass
  • Residual Comparison Test — Interrogates the shape of the leftover residuals — against a null, a rival model, or a raw sample — to tell honest noise from a model that is quietly wrong.
  • Successive-Order Refinement
  • Validity Boundary Scan
  • Zeroth-Order Model Selection

Parameter / Tuning Dimensions

Key tuning dimensions include:

  • how simple the baseline is allowed to be,
  • how close the baseline must be to the target,
  • whether perturbation size is judged quantitatively or qualitatively,
  • how many correction orders are worth adding,
  • what residual error is acceptable,
  • when validation must occur,
  • how local or general the validity range is,
  • when to stop refining and switch to another method,
  • how much interpretability is required, and
  • how high the stakes are if the approximation fails.

Invariants to Preserve

The baseline must be explicitly identified. The perturbation terms must be defined as departures from that baseline. The correction order must be traceable. The residual error must be bounded for the intended use. The validity range must travel with the result. Fallback triggers must exist when the correction logic fails.

The baseline should not become an unquestioned ideology; the perturbation should not become miscellaneous complexity; and the final answer should not hide its approximate status.

Target Outcomes

The expected outcomes are a tractable approximation, clearer reasoning about what is known and what is corrected, faster progression from simple to complex models, better communication of validity boundaries, and disciplined escalation when a perturbative solution is no longer credible.

A successful use of the archetype gives users a result like: “This baseline solution plus these corrections is valid under these parameter ranges and residual-error limits. Outside that range, use a different method.”

Tradeoffs

The main tradeoffs are tractability versus completeness, interpretability versus fidelity, speed versus validation, generality versus local accuracy, and mathematical elegance versus empirical accountability.

A baseline can make reasoning possible, but it can also hide excluded effects. Correction terms can improve realism, but they can also accumulate into an opaque patchwork. The method is powerful because it simplifies; it is risky for the same reason.

Failure Modes

Common failure modes include false baseline correspondence, large perturbations masquerading as small ones, correction pileup, divergent or unstable refinement, validity-range creep, residual blindness, and authority inflation.

The most dangerous failure is treating the exactness of the baseline as if it transfers to the corrected target. The correct communication is not “the target is solved exactly”; it is “the reference case is solved, the target has been approximated through these corrections, and this is the range where that approximation is defensible.”

Neighbor Distinctions

core_model_first begins with a simple core, but does not require explicit perturbation terms or ordered correction logic. bounded_approximation bounds error, but does not prescribe a baseline-plus-correction construction. perturbation_testing disturbs a system to test robustness; this archetype uses perturbations to construct a solution. layered_model_validation validates added layers, while this archetype defines those layers as corrections from a reference solution. hierarchical_decomposition, modular_decomposition, and recursive_problem_decomposition are generic decomposition neighbors; they do not center the solvable-baseline/residual-correction split.

Cross-Domain Examples

In quantum chemistry, a solved atomic model can act as a baseline, with interaction corrections added for harder systems. In climate modeling, an equilibrium energy balance model can be refined with forcing and feedback terms. In aircraft design, a known airfoil can anchor corrections for geometry, loading, and flow regime. In operations planning, a normal schedule can be corrected for seasonal demand, outages, or staffing shortages. In software performance modeling, an ideal service model can be corrected for cache contention, retry storms, queueing, and network jitter.

Across domains, the same structure holds: a reference solution anchors reasoning; departures are named; corrections are ordered; residual error and validity range are communicated.

Non-Examples

A simple estimate with no record of omitted effects is not this archetype. A black-box simulation with every variable included at once is not this archetype. A robustness stress test is not this archetype unless it uses perturbations to construct a corrected solution. A departmental work breakdown is not this archetype unless the partition is specifically between a solved baseline and residual corrections.

Abstractions this archetype builds on — directly (a source ingredient) or as a related pattern. Links follow the typed catalog namespace.

Built directly on (2)

  • Decomposition: Breaking a whole into parts that can be analyzed independently and recombined to reconstitute the whole, making complexity tractable through divide-and-conquer.
  • Perturbation Theory: A technique for handling an intractable problem by splitting it into an exactly solvable baseline plus a small correction, then expanding the quantities of interest as a power series in that small parameter.

Also references 28 related abstractions

Variants

Narrower or domain-specific specializations that share this archetype's core structure. Recognized variants are established; candidate variants are provisional.

First-Order Perturbative Correction · implementation variant · recognized

Use only the leading correction beyond the solvable baseline when higher-order effects are small or unavailable.

  • Distinct from parent: Narrower than the parent because it stops at the first correction rather than managing a full ordered correction sequence.
  • Use when: A baseline solution exists; The first residual term is clearly dominant; Decision value does not justify deeper correction orders.
  • Typical domains: engineering estimation, economics, operations planning
  • Common mechanisms: first order correction pass, residual comparison test

Linear Response Approximation · mechanism family variant · candidate

Approximate a system response by its local linear change around a baseline state.

  • Distinct from parent: More mathematically specific than the parent and may not cover non-linear or non-additive correction methods.
  • Use when: Departures from the baseline are small; Local slope or sensitivity is meaningful; Nonlinear effects can be bounded or monitored.
  • Typical domains: control systems, materials testing, policy sensitivity estimates
  • Common mechanisms: dimensionless small parameter check, validity boundary scan

Baseline-Plus-Residual Modeling · implementation variant · recognized

Represent a phenomenon as a trusted baseline prediction plus a residual correction layer.

  • Distinct from parent: More empirical and data-fitting oriented than the parent’s broader perturbation-theory logic.
  • Use when: A stable baseline forecast or model exists; Residuals show recurring structure; The target output can be recomposed from baseline and residual terms.
  • Typical domains: forecasting, maintenance planning, risk scoring
  • Common mechanisms: delta term isolation, benchmark backtest

Asymptotic Expansion Ladder · scale variant · candidate

Organize corrections as an ordered expansion whose early terms may be useful even when the full sequence does not converge globally.

  • Distinct from parent: More specialized around expansion behavior and stopping rules.
  • Use when: A small parameter or limiting regime exists; Early correction terms improve estimates; Later terms need stopping or truncation rules.
  • Typical domains: physics, applied mathematics, computational modeling
  • Common mechanisms: successive order refinement, convergence or asymptotic behavior check

Benchmark-Anchored Refinement · validation variant · candidate

Use known benchmark cases to decide which correction terms are legitimate and when the baseline-plus-correction model must be revised.

  • Distinct from parent: A validation-heavy subtype of the general archetype.
  • Use when: Exact target solutions are rare but known benchmark cases exist; Correction choices need external validation; Model users need trustable stopping rules.
  • Typical domains: quantum chemistry, aerodynamics, climate modeling
  • Common mechanisms: benchmark backtest, validity boundary scan

Near names: Zeroth-Order Modeling, Baseline-Plus-Delta Analysis, Small-Parameter Approximation, Reference-Case Expansion, Baseline-Residual Decomposition.