Gap MINKT -- definitions #
The decision instance (x, 1^t, 1^s) has the following promised sides:
- yes:
C_U^t(x) <= s; - no:
C_U^(tau(|x|,t))(x) > sigma(|x|,s).
The description transformation sigma and clock transformation tau are
independent parameters. Parameters.IsWidening is kept as a separate property:
it is required to prove the two sides disjoint, but not to state either side or
the associated search relation.
The search relation follows the optimization form: relative to the exact finite
value C_U^t(x) = s, output a program of length at most sigma(|x|,s) that
produces x within tau(|x|,t) steps.
Independent quantitative transformations for description loss and clock blow-up.
Allowed output-program length from input length and source complexity.
Allowed target clock from input length and source clock.
Instances For
Both resources weakly increase. This suffices to make the gap sides disjoint and to reuse an exact source description as a search witness.
Equations
- parameters.IsWidening = ((∀ (length threshold : ℕ), threshold ≤ parameters.description length threshold) ∧ ∀ (length time : ℕ), time ≤ parameters.clock length time)
Instances For
For each output length, increasing the source threshold cannot decrease the allowed target description length.
Equations
- parameters.DescriptionMonotone = ∀ (length : ℕ), Monotone (parameters.description length)
Instances For
The description-loss map is bounded by one polynomial in the sum of the output length and source optimum.
Equations
- parameters.DescriptionPolyBound = ∃ (polynomial : Polynomial ℕ), ∀ (length optimum : ℕ), parameters.description length optimum ≤ Polynomial.eval (length + optimum) polynomial
Instances For
Finite source complexity values are polynomially bounded by the canonical
(x,1^t) input length. This names the precise input-size bridge needed for
optimization-search balance; input locality proves it for every current
sequential-input TM, with the identity polynomial.
Equations
- One or more equations did not get rendered due to their size.
Instances For
Forget the decision threshold, retaining the underlying MINKT instance.
Instances For
Unary encoding of the decision threshold.
Equations
- inst.unaryThreshold = List.replicate inst.threshold true
Instances For
Canonical nested-pair encoding of (x, 1^t, 1^s).
Equations
- inst.encode = Complexity.pair inst.base.encode inst.unaryThreshold
Instances For
Promised no condition
C_U^(tau(|x|,t))(x) > sigma(|x|,s).
Equations
- One or more equations did not get rendered due to their size.
Instances For
A program witnessing the relaxed target resources of a gap instance.
Equations
- One or more equations did not get rendered due to their size.
Instances For
Canonically encoded promised yes language. Malformed codes are outside the promise rather than being assigned to the no side.
Equations
Instances For
Canonically encoded promised no language for the quantitative gap.
Equations
Instances For
Search approximation relative to the exact finite source complexity.
On input (x, 1^t), a related program has length at most
sigma(|x|, C_U^t(x)) and produces x within tau(|x|,t).
Equations
- One or more equations did not get rendered due to their size.
Instances For
Canonical binary-input form of the optimization search relation. Malformed
(x,1^t) codes have no related output.
Equations
- One or more equations did not get rendered due to their size.
Instances For
Semantic search algorithm returning a candidate program from (x,1^t).
Instances For
A search algorithm satisfies the approximation relation on every input whose source time-bounded complexity is finite.
Equations
- One or more equations did not get rendered due to their size.
Instances For
Executably check whether a candidate meets a gap instance's relaxed target resources.
Equations
- One or more equations did not get rendered due to their size.
Instances For
Convert a search algorithm into a total Boolean decision function by decoding a gap instance and checking the returned candidate program.
Equations
- One or more equations did not get rendered due to their size.