class

Crowbar::Evolution::Manager

Inherits Reference < Object

Coordinates genetic algorithm feedback, selection, crossover, Multi-Armed Bandit credit assignment, and anti-stagnation.

Constructors

new(population_size : Int32 = 64)
Source

Instance methods

bandit
Source
corpus
Source
crossover_rate
Source
crossover_rate=(crossover_rate : Float64)
Source
enabled
Source
enabled=(enabled : Bool)
Source
exploration_rate
Source
exploration_rate=(exploration_rate : Float64)
Source
harvest_feedback(text : String) : Array(String)

Extracts tokens, identifiers, and hex constants from target error messages

Source
harvested_tokens
Source
last_applied_mutators
Source
last_applied_mutators=(last_applied_mutators : Array(String))
Source
last_best_fitness
Source
next_parent(baseline : Buffer, prng : PRNG) : Buffer

Selects or generates the starting parent buffer for the next mutation round

Source
report(data : Buffer | Bytes | String, fitness : Float64, feature : String | Nil = nil, coverage_hash : UInt64 | Nil = nil)

Report fitness outcome for a candidate (continuous score) with optional feature & coverage telemetry

Source
report(data : Buffer | Bytes | String, success : Bool, feature : String | Nil = nil, coverage_hash : UInt64 | Nil = nil)

Report binary success outcome (converted to fitness 1.0 vs 0.0)

Source
selection_strategy
Source
selection_strategy=(selection_strategy : Symbol)
Source
stagnation_count
Source
stagnation_limit
Source
stagnation_limit=(stagnation_limit : Int32)
Source
total_evaluations
Source
tournament_size
Source
tournament_size=(tournament_size : Int32)
Source