Class: MachineLearningWorkbench::Optimizer::NaturalEvolutionStrategies::SNES

Inherits:
Base
  • Object
show all
Defined in:
lib/machine_learning_workbench/optimizer/natural_evolution_strategies/snes.rb

Overview

Separable Natural Evolution Strategies

Instance Attribute Summary collapse

Attributes inherited from Base

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

Instance Method Summary collapse

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 Attribute Details

#variances ⇒ Object (readonly)

Returns the value of attribute variances.



6
7
8
 
# File 'lib/machine_learning_workbench/optimizer/natural_evolution_strategies/snes.rb', line 6

def variances
  @variances
end

Instance Method Details

#convergence ⇒ Object

Estimate algorithm convergence as total variance



24
25
26
 
# File 'lib/machine_learning_workbench/optimizer/natural_evolution_strategies/snes.rb', line 24

def convergence
  variances.reduce :+
end

#initialize_distribution(mu_init: 0, sigma_init: 1) ⇒ Object 



8
9
10
11
12
13
 
# File 'lib/machine_learning_workbench/optimizer/natural_evolution_strategies/snes.rb', line 8

def initialize_distribution mu_init: 0, sigma_init: 1
  @mu = NMatrix.new([1, ndims], mu_init, dtype: dtype)
  sigma_init = [sigma_init]*ndims unless sigma_init.kind_of? Enumerable
  @variances = NMatrix.new([1,ndims], sigma_init, dtype: dtype)
  @sigma = NMatrix.diagonal(variances, dtype: dtype)
end

#load(data) ⇒ Object

Raises:

  • (ArgumentError) 


32
33
34
35
36
37
38
 
# File 'lib/machine_learning_workbench/optimizer/natural_evolution_strategies/snes.rb', line 32

def load data
  raise ArgumentError unless data.size == 2
  mu_ary, variances_ary = data
  @mu = NMatrix[*mu_ary, dtype: dtype]
  @variances = NMatrix[*variances_ary, dtype: dtype]
  @sigma = NMatrix.diagonal(variances, dtype: dtype)
end

#save ⇒ Object 



28
29
30
 
# File 'lib/machine_learning_workbench/optimizer/natural_evolution_strategies/snes.rb', line 28

def save
  [mu.to_consistent_a, variances.to_consistent_a]
end

#train(picks: sorted_inds) ⇒ Object 



15
16
17
18
19
20
21
 
# File 'lib/machine_learning_workbench/optimizer/natural_evolution_strategies/snes.rb', line 15

def train picks: sorted_inds
  g_mu = utils.dot(picks)
  g_sigma = utils.dot(picks**2 - 1)
  @mu += sigma.dot(g_mu.transpose).transpose * lrate
  @variances *= (g_sigma * lrate / 2).exponential
  @sigma = NMatrix.diagonal(variances, dtype: dtype)
end