Class: DSPy::Teleprompt::GEPA::ParetoSelector

Inherits:
Object
  • Object
show all
Extended by:
T::Sig
Defined in:
lib/dspy/teleprompt/gepa.rb

Overview

ParetoSelector: Multi-objective optimization using Pareto frontier analysis

Instance Attribute Summary collapse

Instance Method Summary collapse

Constructor Details

#initialize(evaluator:, config:) ⇒ ParetoSelector

Returns a new instance of ParetoSelector.



2480
2481
2482
2483
 
# File 'lib/dspy/teleprompt/gepa.rb', line 2480

def initialize(evaluator:, config:)
  @evaluator = evaluator
  @config = config
end

Instance Attribute Details

#config ⇒ Object (readonly)

Returns the value of attribute config.



2477
2478
2479
 
# File 'lib/dspy/teleprompt/gepa.rb', line 2477

def config
  @config
end

#evaluator ⇒ Object (readonly)

Returns the value of attribute evaluator.



2474
2475
2476
 
# File 'lib/dspy/teleprompt/gepa.rb', line 2474

def evaluator
  @evaluator
end

Instance Method Details

#select_parents(population_with_scores, count:) ⇒ Object 



2487
2488
2489
2490
2491
2492
2493
2494
2495
2496
2497
2498
2499
2500
 
# File 'lib/dspy/teleprompt/gepa.rb', line 2487

def select_parents(population_with_scores, count:)
  return [] if population_with_scores.empty?
  return population_with_scores.map(&:first) if count >= population_with_scores.size
  
  # Combine tournament and Pareto-based selection for parent selection
  selected = []
  
  count.times do
    parent = tournament_selection(population_with_scores)
    selected << parent
  end
  
  selected
end

#select_survivors(population_with_scores, count:) ⇒ Object 



2504
2505
2506
2507
2508
2509
2510
2511
2512
2513
2514
2515
2516
2517
2518
2519
2520
2521
2522
2523
2524
2525
2526
2527
 
# File 'lib/dspy/teleprompt/gepa.rb', line 2504

def select_survivors(population_with_scores, count:)
  return [] if population_with_scores.empty?
  return population_with_scores.map(&:first) if count >= population_with_scores.size
  
  scores = population_with_scores.map(&:last)
  
  # Find Pareto frontier first
  pareto_frontier = find_pareto_frontier(scores)
  frontier_indices = scores.each_index.select { |i| pareto_frontier.include?(scores[i]) }
  frontier_programs = frontier_indices.map { |i| population_with_scores[i].first }
  
  if frontier_programs.size >= count
    # Use diversity selection within frontier
    frontier_with_scores = frontier_indices.map { |i| population_with_scores[i] }
    return diversity_selection(frontier_with_scores, count: count)
  else
    # Include all frontier + fill remaining with elite selection
    remaining_count = count - frontier_programs.size
    remaining_population = population_with_scores.reject.with_index { |_, i| frontier_indices.include?(i) }
    
    additional = elite_selection(remaining_population, count: remaining_count)
    frontier_programs + additional
  end
end