Problem#

Layer 1: Use

`saealib.Problem`	Definition of optimization problem.
`saealib.PymooProblem`	Adapter wrapping a pymoo Problem instance as a saealib `Problem`.
`saealib.InequalityConstraint`	A single inequality constraint: g(x) <= threshold.
`saealib.EqualityConstraint`	A single equality constraint: h(x) = 0, satisfied within a tolerance.
`saealib.ConstraintHandler`	Pluggable constraint-processing strategy.
`saealib.StaticToleranceHandler`	Default constraint handler reproducing the static-tolerance behavior.
`saealib.EpsilonConstraintHandler`	Dynamic ε-constraint handler.
`saealib.GradientRepairHandler`	Gradient-based constraint repair handler.

saealib.linear_epsilon_schedule(eps0, n_gen)[source]#

Return a linear ε schedule: eps0 at gen 0, 0.0 at gen n_gen.

Parameters:

eps0 (float) – Initial ε value (gen 0).
n_gen (int) – Generation at which ε reaches 0.

Returns:

schedule(gen) -> float

Return type:

callable

saealib.exponential_epsilon_schedule(eps0, decay)[source]#

Return an exponential ε schedule: eps0 * decay**gen.

Parameters:

eps0 (float) – Initial ε value (gen 0).
decay (float) – Multiplicative decay factor per generation (0 < decay < 1).

Returns:

schedule(gen) -> float

Return type:

callable

Problem accepts a SearchSpace through space and can manage Genome representations and conversion to evaluation inputs. See Problem and SearchSpace for details.