Class: MachineLearningWorkbench::Optimizer::NaturalEvolutionStrategies::FNES

Inherits:
RNES
  • Object
show all
Defined in:
lib/machine_learning_workbench/optimizer/natural_evolution_strategies/fnes.rb

Overview

Fixed Variance Natural Evolution Strategies

Instance Attribute Summary

Attributes inherited from RNES

#variance

Attributes inherited from Base

#best, #eye, #last_fits, #mu, #ndims, #obj_fn, #opt_type, #parallel_fit, #rescale_lrate, #rescale_popsize, #rng, #sigma

Instance Method Summary collapse

Methods inherited from RNES

#convergence, #initialize_distribution, #load, #save

Methods inherited from Base

#cmaes_lrate, #cmaes_popsize, #cmaes_utilities, #initialize, #interface_methods, #lrate, #move_inds, #popsize, #sorted_inds, #standard_normal_sample, #standard_normal_samples, #utils

Constructor Details

This class inherits a constructor from MachineLearningWorkbench::Optimizer::NaturalEvolutionStrategies::Base

Instance Method Details

#train(picks: sorted_inds) ⇒ Object 



7
8
9
10
 
# File 'lib/machine_learning_workbench/optimizer/natural_evolution_strategies/fnes.rb', line 7

def train picks: sorted_inds
  g_mu = utils.dot(picks)
  @mu += sigma.dot(g_mu.transpose).transpose * lrate
end