Class: Musa::Darwin::Darwin

Inherits:
Object show all
Defined in:
lib/musa-dsl/generative/darwin.rb

Overview

Evolutionary selector for population-based optimization.

Evaluates population using measures and weights, returning sorted population by fitness score.

Defined Under Namespace

Classes: MainContext, Measure, MeasuresEvalContext

Instance Method Summary collapse

Constructor Details

#initialize { ... } ⇒ void

Creates Darwin selector with evaluation rules.

Examples:

darwin = Darwin.new do
  measures { |obj| dimension :value, obj[:score] }
  weight value: 1.0
end

Yields:

  • evaluation DSL block

Yield Returns:

  • (void)

Raises:

  • (ArgumentError) —

    if no block given



109
110
111
112
113
114
115
116
 
# File 'lib/musa-dsl/generative/darwin.rb', line 109

def initialize(&block)
  raise ArgumentError, 'block is needed' unless block

  main_context = MainContext.new &block

  @measures = main_context._measures
  @weights = main_context._weights
end

Instance Method Details

#evaluate_weights(measure_a, measure_b) ⇒ Integer

This method is part of a private API. You should avoid using this method if possible, as it may be removed or be changed in the future.

Compares two measures by their weighted fitness.

Parameters:

  • measure_a (Measure) —

    first measure to compare

  • measure_b (Measure) —

    second measure to compare

Returns:

  • (Integer) —

    comparison result (-1, 0, 1)



188
189
190
 
# File 'lib/musa-dsl/generative/darwin.rb', line 188

def evaluate_weights(measure_a, measure_b)
  measure_b.evaluate_weight(@weights) <=> measure_a.evaluate_weight(@weights)
end

#select(population) ⇒ Array

Selects and ranks population by fitness.

Evaluates each object with measures, normalizes dimensions across population, applies weights, and returns population sorted by fitness (highest first). Objects marked as died are excluded.

Examples:

candidates = [obj1, obj2, obj3, ...]
ranked = darwin.select(candidates)
best = ranked.first      # Highest fitness
worst = ranked.last      # Lowest fitness
top10 = ranked.first(10) # Top 10

Parameters:

  • population (Array) —

    objects to evaluate

Returns:

  • (Array) —

    population sorted by fitness (descending)



134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
 
# File 'lib/musa-dsl/generative/darwin.rb', line 134

def select(population)
  measured_objects = []

  population.each do |object|
    context = MeasuresEvalContext.new

    context.with object, **{}, &@measures
    measure = context._measure

    measured_objects << { object: object, measure: context._measure } unless measure.died?
  end

  limits = {}

  measured_objects.each do |measured_object|
    measure = measured_object[:measure]

    measure.dimensions.each do |measure_name, value|
      limit = limits[measure_name] ||= { min: nil, max: nil }

      limit[:min] = value.to_f if limit[:min].nil? || limit[:min] > value
      limit[:max] = value.to_f if limit[:max].nil? || limit[:max] < value

      limit[:range] = limit[:max] - limit[:min]
    end
  end

  # warn "Darwin.select: weights #{@weights}"

  measured_objects.each do |measured_object|
    measure = measured_object[:measure]

    measure.dimensions.each do |dimension_name, value|
      limit = limits[dimension_name]
      measure.normalized_dimensions[dimension_name] =
        limit[:range].zero? ? 0.5 : (value - limit[:min]) / limit[:range]
    end

    # warn "Darwin.select: #{measured_object[:object]} #{measured_object[:measure]} weight=#{measured_object[:measure].evaluate_weight(@weights).round(2)}"
  end

  measured_objects.sort! { |a, b| evaluate_weights a[:measure], b[:measure] }

  measured_objects.collect { |measured_object| measured_object[:object] }
end