machine_learning_workbench 0.4.0 → 0.4.2

checksums.yaml

@@ -1,7 +1,7 @@
 ---
 SHA1:
-  metadata.gz: 743cc4c65bda521785e00dc563b20fcd5660a6af
-  data.tar.gz: 80924d34dc550df7b25d565022f50297b8bb72d5
+  metadata.gz: 05de54fb5221e0b9c4e0a7f59f79ed3b0fb7c166
+  data.tar.gz: da580d26078824e27aff0a518d8a57080d06c2d7
 SHA512:
-  metadata.gz: 2ef96c2884a6f43304f0dba63239a7469b80c7be46c97af7c8de283a448a8ee049b55cdc98b0aa5bfde82fc024fde891c982ec621bb4cafcd6070814e8366976
-  data.tar.gz: 0b5571eb26babf90015deec8425dc0a9c75bebd47ef50b4927a131e86f5f0c9e6e828e3ee1b212597222b9aa7209cd54f6b3db7a9c8ec6afcceafe94016165a1
+  metadata.gz: 50df1af8dc0cfdf8c3ecf3e99553749f25fc562ab022032b31d43bbc7986f17ec2a444a1a60efb06c55f2c85a825986fd45d3f8753ff3eb81b9f51549603e476
+  data.tar.gz: c84df475c3be8175d7c46c83bfdc755803d63c060f1ed7ae4749602003a7f4917726a89412d63011b9d144d241cbf35c24c1cfd871beeae5d379cada304aba23

data/.travis.yml

@@ -5,5 +5,5 @@ rvm:
 addons:
   apt:
     packages:
-    - libatlas-base-dev
-before_install: gem install bundler -v 1.16.0
+    - libopenblas-base
+before_install: gem install bundler -v 1.16.0

data/lib/machine_learning_workbench/neural_network/base.rb

@@ -44,16 +44,15 @@ module MachineLearningWorkbench::NeuralNetwork
     def reset_state
       state.each do |s|
         s.fill 0           # reset state to zero
-        s[0,-1] = 1        # add bias
+        s[-1] = 1        # add bias
       end
-      state[-1][0,-1] = 0  # last layer has no bias
+      state[-1][-1] = 0  # last layer has no bias
     end
 
     # Initialize the network with random weights
     def init_random
-      # Will only be used for testing, no sense optimizing it now (NArray#rand)
       # Reusing `#load_weights` instead helps catching bugs
-      load_weights nweights.times.collect { rand(-1.0..1.0) }
+      load_weights NArray.new(nweights).rand(-1,1)
     end
 
     ## Weight utilities
@@ -88,10 +87,9 @@ module MachineLearningWorkbench::NeuralNetwork
     end
 
     # Returns the weight matrix
-    # @return [Array] three-dimensional Array of weights: a list of weight
-    #   matrices, one for each layer.
+    # @return [Array<NArray>] list of NArray matrices of weights (one per layer).
     def weights
-      layers.collect(&:to_a)
+      layers
     end
 
     # Number of neurons per layer. Although this implementation includes inputs
@@ -126,12 +124,13 @@ module MachineLearningWorkbench::NeuralNetwork
     #   all goes well there's nothing to return but a confirmation to the caller.
     def load_weights weights
       raise ArgumentError unless weights.size == nweights
-      weights_iter = weights.each
-      @layers ||= layer_shapes.collect { |shape| NArray.zeros shape }
-      layers.each do |narr|
-        narr.each_with_index do |_val, *idxs|
-          narr[*idxs] = weights_iter.next
-        end
+      weights = weights.to_na unless weights.kind_of? NArray
+      from = 0
+      @layers = layer_shapes.collect do |shape|
+        to = from + shape.reduce(:*)
+        lay_w = weights[from...to].reshape *shape
+        from = to
+        lay_w
       end
       reset_state
       return true
@@ -145,52 +144,49 @@ module MachineLearningWorkbench::NeuralNetwork
     # @return [Array] the activation of the output layer
     def activate input
       raise ArgumentError unless input.size == struct.first
-      raise ArgumentError unless input.is_a? Array
       # load input in first state
-      @state[0][0, 0..-2] = input
+      state[0][0...struct.first] = input
       # activate layers in sequence
       nlayers.times.each do |i|
         act = activate_layer i
-        @state[i+1][0, 0...act.size] = act
+        state[i+1][0...act.size] = act
       end
       return out
     end
 
     # Extract and convert the output layer's activation
-    # @return [Array] the activation of the output layer as 1-dim Array
+    # @return [NArray] the activation of the output layer
     def out
-      state.last.to_a.flatten
+      state.last
     end
 
-    # define #activate_layer in child class
-
     ## Activation functions
 
     # Traditional sigmoid with variable steepness
     def sigmoid k=0.5
       # k is steepness:  0<k<1 is flatter, 1<k is flatter
       # flatter makes activation less sensitive, better with large number of inputs
-      lambda { |x| 1.0 / (Numo::NMath.exp(-k * x) + 1.0) }
+      -> (x) { 1.0 / (Numo::NMath.exp(-k * x) + 1.0) }
     end
 
     # Traditional logistic
     def logistic
-      lambda { |x|
+      -> (x) do
         exp = Numo::NMath.exp(x)
         # exp.infinite? ? exp : exp / (1.0 + exp)
         exp / (1.0 + exp)
-      }
+      end
     end
 
     # LeCun hyperbolic activation
     # @see http://yann.lecun.com/exdb/publis/pdf/lecun-98b.pdf Section 4.4
     def lecun_hyperbolic
-      lambda { |x| 1.7159 * Numo::NMath.tanh(2.0*x/3.0) + 1e-3*x }
+      -> (x) { 1.7159 * Numo::NMath.tanh(2.0*x/3.0) + 1e-3*x }
     end
 
     # Rectified Linear Unit (ReLU)
     def relu
-      lambda { |x| (x>0).all? && x || x.class.zeros(x.shape) }
+      -> (x) { (x>0).all? && x || x.class.zeros(x.shape) }
     end
 
 

data/lib/machine_learning_workbench/neural_network/recurrent.rb

@@ -14,21 +14,29 @@ module MachineLearningWorkbench::NeuralNetwork
       end
     end
 
+      # # NOTE: current layer index corresponds to index of next state!
+      # previous = nlay     # index of previous layer (inputs)
+      # current = nlay + 1  # index of current layer (outputs)
+      # # Copy the level's last-time activation to the input (previous state)
+      # # TODO: ranges in `NArray#[]` should be reliable, get rid of loop
+      # nneurs(current).times do |i| # for each activations to copy
+      #   # Copy output from last-time activation to recurrency in previous state
+      #   @state[previous][0, nneurs(previous) + i] = state[current][0, i]
+      # end
+      # act_fn.call state[previous].dot layers[nlay]
+
     # Activates a layer of the network.
     # Bit more complex since it has to copy the layer's activation on
     # last input to its own inputs, for recursion.
     # @param i [Integer] the layer to activate, zero-indexed
-    def activate_layer nlay #_layer
-      # NOTE: current layer index corresponds to index of next state!
-      previous = nlay     # index of previous layer (inputs)
-      current = nlay + 1  # index of current layer (outputs)
-      # Copy the level's last-time activation to the input (previous state)
-      # TODO: ranges in `NArray#[]` should be reliable, get rid of loop
-      nneurs(current).times do |i| # for each activations to copy
-        # Copy output from last-time activation to recurrency in previous state
-        @state[previous][0, nneurs(previous) + i] = state[current][0, i]
-      end
-      act_fn.call state[previous].dot layers[nlay]
+    def activate_layer nlay
+      # Mark begin and end of recursion outputs in current state
+      begin_recur = nneurs(nlay)
+      end_recur = nneurs(nlay) + nneurs(nlay+1)
+      # Copy the level's last-time activation to the current input recurrency
+      state[nlay][begin_recur...end_recur] = state[nlay+1][0...nneurs(nlay+1)]
+      # Activate current layer
+      act_fn.call state[nlay].dot layers[nlay]
     end
 
   end

data/lib/machine_learning_workbench/optimizer/natural_evolution_strategies/base.rb

@@ -119,7 +119,7 @@ module MachineLearningWorkbench::Optimizer::NaturalEvolutionStrategies
 
       sort_idxs = fits.sort_index
       sort_idxs = sort_idxs.reverse if opt_type == :min
-      this_best = [fits[sort_idxs[-1]], inds[sort_idxs[-1]]]
+      this_best = [fits[sort_idxs[-1]], inds[sort_idxs[-1], true]]
 
       opt_cmp_fn = opt_type==:min ? :< : :>
       @best = this_best if this_best.first.send(opt_cmp_fn, best.first)

data/lib/machine_learning_workbench/optimizer/natural_evolution_strategies/bdnes.rb

@@ -5,8 +5,7 @@ module MachineLearningWorkbench::Optimizer::NaturalEvolutionStrategies
 
     MAX_RSEED = 10**Random.new_seed.size # same range as Random.new_seed
 
-    attr_reader :ndims_lst, :obj_fn, :opt_type, :parallel_fit, :blocks, :popsize, :rng,
-      :best, :last_fits
+    attr_reader :ndims_lst, :blocks, :popsize
 
     # Initialize a list of XNES, one for each block
     # see class `Base` for the description of the rest of the arguments.
@@ -67,7 +66,7 @@ module MachineLearningWorkbench::Optimizer::NaturalEvolutionStrategies
       # sorted_samples = sorted.map(&:last)
       sort_idxs = fits.sort_index
       sort_idxs = sort_idxs.reverse if opt_type == :min
-      this_best = [fits[sort_idxs[-1]], full_inds[sort_idxs[-1]]]
+      this_best = [fits[sort_idxs[-1]], full_inds[sort_idxs[-1], true]]
       opt_cmp_fn = opt_type==:min ? :< : :>
       @best = this_best if this_best.first.send(opt_cmp_fn, best.first)
       sorted_samples = full_samples.values_at *sort_idxs

data/machine_learning_workbench.gemspec

@@ -34,21 +34,18 @@ Gem::Specification.new do |spec|
 
   # Test
   spec.add_development_dependency "rspec", "~> 3.0"
-  spec.add_development_dependency "rmagick"  # only used for one example
+  spec.add_development_dependency "rmagick"  # uhm would gladly drop this
 
   # Debug
   spec.add_development_dependency "pry", "~> 0.10"
   spec.add_development_dependency "pry-nav", "~> 0.2"
   spec.add_development_dependency "pry-rescue", "~> 1.4"
   spec.add_development_dependency "pry-stack_explorer", "~> 0.4"
+  spec.add_development_dependency "pry-doc", "~> 0.12"
 
   # Run
   spec.requirements << "libopenblas-base"  # library for following dependency
-  spec.add_dependency "numo-linalg"
+  spec.add_dependency "numo-narray", "~> 0.9"
+  spec.add_dependency "numo-linalg", "~> 0.1"
   spec.add_dependency "parallel", "~> 1.12"
-
-
-
-  # DELETEME
-  spec.add_dependency "nmatrix-atlas"
 end

metadata

@@ -1,14 +1,14 @@
 --- !ruby/object:Gem::Specification
 name: machine_learning_workbench
 version: !ruby/object:Gem::Version
-  version: 0.4.0
+  version: 0.4.2
 platform: ruby
 authors:
 - Giuseppe Cuccu
 autorequire: 
 bindir: bin
 cert_chain: []
-date: 2018-03-25 00:00:00.000000000 Z
+date: 2018-03-26 00:00:00.000000000 Z
 dependencies:
 - !ruby/object:Gem::Dependency
   name: bundler
@@ -123,47 +123,61 @@ dependencies:
       - !ruby/object:Gem::Version
         version: '0.4'
 - !ruby/object:Gem::Dependency
-  name: numo-linalg
+  name: pry-doc
   requirement: !ruby/object:Gem::Requirement
     requirements:
-    - - ">="
+    - - "~>"
       - !ruby/object:Gem::Version
-        version: '0'
+        version: '0.12'
+  type: :development
+  prerelease: false
+  version_requirements: !ruby/object:Gem::Requirement
+    requirements:
+    - - "~>"
+      - !ruby/object:Gem::Version
+        version: '0.12'
+- !ruby/object:Gem::Dependency
+  name: numo-narray
+  requirement: !ruby/object:Gem::Requirement
+    requirements:
+    - - "~>"
+      - !ruby/object:Gem::Version
+        version: '0.9'
   type: :runtime
   prerelease: false
   version_requirements: !ruby/object:Gem::Requirement
     requirements:
-    - - ">="
+    - - "~>"
       - !ruby/object:Gem::Version
-        version: '0'
+        version: '0.9'
 - !ruby/object:Gem::Dependency
-  name: parallel
+  name: numo-linalg
   requirement: !ruby/object:Gem::Requirement
     requirements:
     - - "~>"
       - !ruby/object:Gem::Version
-        version: '1.12'
+        version: '0.1'
   type: :runtime
   prerelease: false
   version_requirements: !ruby/object:Gem::Requirement
     requirements:
     - - "~>"
       - !ruby/object:Gem::Version
-        version: '1.12'
+        version: '0.1'
 - !ruby/object:Gem::Dependency
-  name: nmatrix-atlas
+  name: parallel
   requirement: !ruby/object:Gem::Requirement
     requirements:
-    - - ">="
+    - - "~>"
       - !ruby/object:Gem::Version
-        version: '0'
+        version: '1.12'
   type: :runtime
   prerelease: false
   version_requirements: !ruby/object:Gem::Requirement
     requirements:
-    - - ">="
+    - - "~>"
       - !ruby/object:Gem::Version
-        version: '0'
+        version: '1.12'
 description: |-
   This workbench holds a collection of machine learning
   methods in Ruby. Rather than specializing on a single task or method, this