Reference · The Applied Higher-Order Functions catalogue

rainbow.algorithms.proofs.ahof.surrogate-explanation

The Applied Higher-Order Functions cataloguesrc/rainbow/algorithms/proofs/ahof/surrogate_explanation.clj1 definitions, 4 theorems

AHoF V56 Surrogate Model Explanation: fit an interpretable surrogate to a black-box model and explain via the surrogate; the explanation is faithful when the surrogate's local fidelity error is within the tolerance: faithful? error tolerance = decide (error <= tolerance). Proved native and axiom-free (decision, two positive instances, at-most certificate, negative instance). Requires audit/decide-nnrat-at-most, nnrat/at-most, nnrat/decide-at-most-true-sound (all in the D63 checkpoint). Scope (measured by the kernel): this module has 4 worked instances (statements about specific values). None of its theorems is a guarantee about the algorithm for all inputs; `lein verify rainbow.algorithms.proofs.ahof.surrogate-explanation` prints this classification.

surrogate-exp/faithful? :
  (x₁ x₂ : ℚ≥0)
  → data/Bool

surrogate-exp/low-error-faithful :
  → lis/extends? 1/4 1/2 = data/true

surrogate-exp/boundary-faithful :
  → lis/extends? 1/2 1/2 = data/true

surrogate-exp/fidelity-certificate :
  → 1/4 ≤ 1/2

surrogate-exp/high-error-unfaithful :
  → lis/extends? 3/4 1/2 = data/false