Behaviour for fitness evaluation functions.
Fitness functions determine how well an entity performs against the optimization criteria.
Summary
Callbacks
Batch evaluate multiple entities for efficiency.
Compare two entities directly without computing scores.
Evaluate a single entity's fitness.
Functions
Shared batch evaluation implementation.
Extract score from evaluate/2 result, raising on invalid format.
Types
Callbacks
Batch evaluate multiple entities for efficiency.
Default implementation calls evaluate/2 for each entity. Override for performance when batch evaluation is available.
Compare two entities directly without computing scores.
This can be more efficient than scoring both entities when only relative fitness matters.
@callback evaluate(entity(), context()) :: eval_result()
Evaluate a single entity's fitness.
Returns either a simple score or a score with metadata. Higher scores indicate better fitness.
Examples
def evaluate(text, _context) do
similarity = String.jaro_distance(text, @target)
{:ok, similarity}
end
def evaluate(text, _context) do
score = String.jaro_distance(text, @target)
{:ok, %{score: score, metadata: %{length: String.length(text)}}}
endFunctions
Shared batch evaluation implementation.
Evaluates multiple entities using the provided module's evaluate/2 callback.
@spec score_or_raise!(eval_result()) :: score()
Extract score from evaluate/2 result, raising on invalid format.
Handles both simple scores and score maps with optional metadata.