neuro 0.4.0

data/tests/runner.rb

@@ -0,0 +1,18 @@
+#!/usr/bin/env ruby
+
+require 'test/unit/ui/console/testrunner'
+require 'test/unit/testsuite'
+$:.unshift File.expand_path(File.dirname($0))
+require 'test_parity'
+require 'test_even_odd'
+
+class TS_AllTests
+  def self.suite
+    suite = Test::Unit::TestSuite.new
+    suite << TC_Parity.suite
+    suite << TC_EvenOdd.suite
+    suite
+  end
+end
+Test::Unit::UI::Console::TestRunner.run(TS_AllTests)
+  # vim: set et sw=2 ts=2:

data/tests/test_even_odd.rb

@@ -0,0 +1,69 @@
+#!/usr/bin/env ruby
+#
+# Stupid little test, that finds out if 5-bit numbers are even or odd.
+#
+require 'test/unit'
+require 'neuro'
+
+class TC_EvenOdd < Test::Unit::TestCase
+  include Neuro
+
+  @@even = [
+   [ 0, 0, 1, 1, 0 ],
+   [ 1, 1, 0, 1, 0 ],
+   [ 1, 0, 1, 1, 0 ],
+   [ 0, 1, 0, 1, 0 ],
+   [ 1, 0, 1, 0, 0 ],
+   [ 0, 1, 0, 1, 0 ],
+  ]
+
+  @@odd = [
+   [ 1, 0, 0, 1, 1 ],
+   [ 1, 1, 1, 0, 1 ],
+   [ 0, 0, 0, 1, 1 ],
+   [ 0, 1, 0, 0, 1 ],
+   [ 0, 1, 0, 1, 1 ],
+   [ 1, 1, 1, 1, 1 ],
+  ]
+
+  def setup
+    @eta = 0.2
+    filename = 'test_even_odd.rb.dump'
+    if File.exist?(filename)
+      File.open(filename) do |f|
+        @network = Network.load(f)
+      end
+    else
+      @network = Network.new(5, 2, 1)
+      @network.debug = STDERR
+      @network.debug_step = 1000
+      max_error = 1.0E-5
+      max_count = (@@even.size + @@odd.size) * 10
+      count = max_count
+      until count < max_count
+        count = 0
+        desired = [ 0.9 ]
+        @@even.each do |sample|
+          count += @network.learn(sample, desired, max_error, @eta)
+        end
+        desired = [ 0.1 ]
+        @@odd.each do |sample|
+          count += @network.learn(sample, desired, max_error, @eta)
+        end
+      end
+      File.open(filename, 'wb') do |f|
+        @network.dump(f)
+      end
+    end
+  end
+
+  def test_hypothesis
+    for x in 0..31
+      vector = (0..4).map { |i| x[i] }.reverse
+      result, = @network.decide vector
+      bit = (result - 0.9).abs < @eta ? 0 : 1
+      assert_equal x[0], bit
+    end
+  end
+end
+  # vim: set et sw=2 ts=2:

data/tests/test_parity.rb

@@ -0,0 +1,58 @@
+#!/usr/bin/env ruby
+
+require 'test/unit'
+require 'neuro'
+
+class TC_Parity < Test::Unit::TestCase
+  include Neuro
+
+  MAX_BITS = 4
+  def setup
+    filename = 'test_parity.rb.dump'
+    if File.exist?(filename)
+      File.open(filename) do |f|
+         @network = Network.load(f)
+      end
+    else
+      @network = Network.new(MAX_BITS, MAX_BITS * 2, 1)
+      @network.debug = STDERR
+      @network.debug_step = 1000
+      eta = 0.2
+      max_error = 1.0E-6
+      vectors = all_vectors
+      max_count = vectors.size * 10
+      count = max_count
+      until count < max_count
+        count = 0
+        vectors.sort_by { rand }.each do |sample|
+          desired = [ parity(sample) == 1 ? 0.9 : 0.1 ]
+          count += @network.learn(sample, desired, max_error, eta)
+        end
+      end
+      File.open(filename, 'wb') do |f|
+        @network.dump(f)
+      end
+    end
+  end
+
+  def parity(vector)
+    (vector.inject(1) { |s,x| s * (x < 0.5 ? -1 : 1) }) < 0 ? 1 : 0
+  end
+
+  def all_vectors
+    vectors = []
+    for x in 0...(2 ** MAX_BITS)
+      vectors << (0...MAX_BITS).map { |i| x[i].zero? ? 0.0 : 1.0 }.reverse
+    end
+    vectors
+  end
+
+  def test_parities
+    all_vectors.each do |vector|
+      result, = @network.decide vector
+      result_parity = result > ((0.9 - 0.1) / 2) ? 1 : 0
+      assert_equal parity(vector), result_parity
+    end
+  end
+end
+  # vim: set et sw=2 ts=2:

metadata

@@ -0,0 +1,59 @@
+--- !ruby/object:Gem::Specification 
+rubygems_version: 0.8.11
+specification_version: 1
+name: neuro
+version: !ruby/object:Gem::Version 
+  version: 0.4.0
+date: 2005-10-03 00:00:00 +02:00
+summary: Neural Network Extension for Ruby
+require_paths: 
+  - ext
+email: flori@ping.de
+homepage: http://neuro.rubyforge.org
+rubyforge_project: neuro
+description: "A Ruby extension that provides a 2-Layer Back Propagation Neural Network, which
+  can be used to categorize datasets of arbitrary size."
+autorequire: neuro
+default_executable: 
+bindir: bin
+has_rdoc: true
+required_ruby_version: !ruby/object:Gem::Version::Requirement 
+  requirements: 
+    - 
+      - ">"
+      - !ruby/object:Gem::Version 
+        version: 0.0.0
+  version: 
+platform: ruby
+signing_key: 
+cert_chain: 
+authors: 
+  - Florian Frank
+files: 
+  - examples
+  - Rakefile
+  - VERSION
+  - install.rb
+  - ext
+  - lib
+  - tests
+  - examples/ocr.rb
+  - ext/extconf.rb
+  - ext/neuro.c
+  - lib/neuro
+  - lib/neuro/display.rb
+  - tests/runner.rb
+  - tests/test_even_odd.rb
+  - tests/test_parity.rb
+test_files: 
+  - tests/runner.rb
+rdoc_options: 
+  - "--main"
+  - Neuro
+extra_rdoc_files: 
+  - ext/neuro.c
+executables: []
+extensions: 
+  - ext/extconf.rb
+requirements: []
+dependencies: []