Class: MachineLearningWorkbench::Optimizer::NaturalEvolutionStrategies::RNES

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

Overview

Radial Natural Evolution Strategies

Direct Known Subclasses

FNES

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

#variance ⇒ Object (readonly)

Returns the value of attribute variance.



5
6
7
 
# File 'lib/machine_learning_workbench/optimizer/natural_evolution_strategies/rnes.rb', line 5

def variance
  @variance
end

Instance Method Details

#convergence ⇒ Object

Estimate algorithm convergence based on variance



23
24
25
 
# File 'lib/machine_learning_workbench/optimizer/natural_evolution_strategies/rnes.rb', line 23

def convergence
  variance
end

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

Raises:

  • (ArgumentError) 


7
8
9
10
11
12
 
# File 'lib/machine_learning_workbench/optimizer/natural_evolution_strategies/rnes.rb', line 7

def initialize_distribution mu_init: 0, sigma_init: 1
  @mu = NMatrix.new([1, ndims], mu_init, dtype: dtype)
  raise ArgumentError unless sigma_init.kind_of? Numeric
  @variance = sigma_init
  @sigma = id * variance
end

#load(data) ⇒ Object

Raises:

  • (ArgumentError) 


31
32
33
34
35
36
 
# File 'lib/machine_learning_workbench/optimizer/natural_evolution_strategies/rnes.rb', line 31

def load data
  raise ArgumentError unless data.size == 2
  mu_ary, @variance = data
  @mu = NMatrix[*mu_ary, dtype: dtype]
  @sigma = id * variance
end

#save ⇒ Object 



27
28
29
 
# File 'lib/machine_learning_workbench/optimizer/natural_evolution_strategies/rnes.rb', line 27

def save
  [mu.to_consistent_a, variance]
end

#train(picks: sorted_inds) ⇒ Object 



14
15
16
17
18
19
20
 
# File 'lib/machine_learning_workbench/optimizer/natural_evolution_strategies/rnes.rb', line 14

def train picks: sorted_inds
  g_mu = utils.dot(picks)
  g_sigma = utils.dot(picks.row_norms**2 - ndims).first # back to scalar
  @mu += sigma.dot(g_mu.transpose).transpose * lrate
  @variance *= Math.exp(g_sigma * lrate / 2)
  @sigma = id * variance
end