hopfield 1.0
This diff represents the content of publicly available package versions that have been released to one of the supported registries. The information contained in this diff is provided for informational purposes only and reflects changes between package versions as they appear in their respective public registries.
- data/LICENSE +22 -0
- data/README.md +30 -0
- data/lib/hopfield.rb +10 -0
- data/lib/hopfield/network.rb +92 -0
- data/lib/hopfield/neuron.rb +18 -0
- data/lib/hopfield/training.rb +47 -0
- metadata +85 -0
data/LICENSE
ADDED
@@ -0,0 +1,22 @@
|
|
1
|
+
Copyright (c) 2012 Bart Olsthoorn, website: bartolsthoorn.nl
|
2
|
+
|
3
|
+
Permission is hereby granted, free of charge, to any person
|
4
|
+
obtaining a copy of this software and associated documentation
|
5
|
+
files (the "Software"), to deal in the Software without
|
6
|
+
restriction, including without limitation the rights to use,
|
7
|
+
copy, modify, merge, publish, distribute, sublicense, and/or sell
|
8
|
+
copies of the Software, and to permit persons to whom the
|
9
|
+
Software is furnished to do so, subject to the following
|
10
|
+
conditions:
|
11
|
+
|
12
|
+
The above copyright notice and this permission notice shall be
|
13
|
+
included in all copies or substantial portions of the Software.
|
14
|
+
|
15
|
+
THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
|
16
|
+
EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES
|
17
|
+
OF MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND
|
18
|
+
NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT
|
19
|
+
HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY,
|
20
|
+
WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING
|
21
|
+
FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR
|
22
|
+
OTHER DEALINGS IN THE SOFTWARE.
|
data/README.md
ADDED
@@ -0,0 +1,30 @@
|
|
1
|
+
# Hopfield Network in Ruby
|
2
|
+
|
3
|
+
## What is it?
|
4
|
+
[Hopfield Networks](http://en.wikipedia.org/wiki/Hopfield_network) model the way humans recall memories, or more specific, how neurons recall the pattern. This means you first train the network with a set of known patterns and then pass an unknown or perturbed version of the pattern. The neurons will restore the missing information to create an exact match.
|
5
|
+
|
6
|
+
The patterns can be passed using multi dimensional array of either 0 and 1 or -1 and 1. An artifical neural network will learn the patterns. Now let's move on to an example.
|
7
|
+
|
8
|
+
## How do I use it?
|
9
|
+
```ruby
|
10
|
+
training = Hopfield::Training.new([pattern1, pattern2])
|
11
|
+
network = Hopfield::Network.new(training, perturbed_pattern)
|
12
|
+
network.pattern # the matched pattern
|
13
|
+
network.runs # how many propagations it took
|
14
|
+
```
|
15
|
+
|
16
|
+
## Example with images
|
17
|
+
See examples/image.rb for a memory association of Charlie Sheen, with a cat hiding him.
|
18
|
+
```
|
19
|
+
$ cd examples
|
20
|
+
$ ruby image.rb
|
21
|
+
Image 1 is now in an array of [20x20]
|
22
|
+
Image 2 is now in an array of [20x20]
|
23
|
+
Hopfield neurons are trained!
|
24
|
+
Neurons propagated: 1776
|
25
|
+
Errors: [0]
|
26
|
+
```
|
27
|
+
The script also creates black and white pattern images for you.
|
28
|
+
|
29
|
+
## Credits
|
30
|
+
I was introduced to Hopfield networks through the book [Clever Algorithms](www.cleveralgorithms.com), and I've borrowed bits of the implementation shown in the book. Also used the `.associated?` syntax found here: [Brain](https://github.com/brainopia/brain).
|
data/lib/hopfield.rb
ADDED
@@ -0,0 +1,10 @@
|
|
1
|
+
require_relative 'hopfield/neuron'
|
2
|
+
require_relative 'hopfield/training'
|
3
|
+
require_relative 'hopfield/network'
|
4
|
+
|
5
|
+
module Hopfield
|
6
|
+
# Hopfield consists of three parts:
|
7
|
+
# => Neuron (hopfield/neuron.rb)
|
8
|
+
# => Training (hopfield/training.rb)
|
9
|
+
# => Network (hopfield/network.rb)
|
10
|
+
end
|
@@ -0,0 +1,92 @@
|
|
1
|
+
module Hopfield
|
2
|
+
class Network
|
3
|
+
attr_accessor :neurons, :patterns, :state, :pattern_width, :vector, :last_error, :runs
|
4
|
+
|
5
|
+
def initialize(training, perturbed_pattern)
|
6
|
+
unless training.class.to_s == 'Hopfield::Training'
|
7
|
+
raise TypeError, 'Training has to be an instance of Hopfield::Training'
|
8
|
+
end
|
9
|
+
|
10
|
+
unless training.patterns.first.size == perturbed_pattern.size
|
11
|
+
raise SyntaxError, 'Given pattern does not match size of the training patterns'
|
12
|
+
end
|
13
|
+
|
14
|
+
# Turn 0 into -1
|
15
|
+
perturbed_pattern.map {|value| (value == 0 ? -1 : value) }
|
16
|
+
|
17
|
+
self.neurons = training.neurons
|
18
|
+
self.patterns = training.patterns
|
19
|
+
self.pattern_width = training.pattern_width
|
20
|
+
self.vector = perturbed_pattern.flatten
|
21
|
+
self.neurons.each_with_index { |neuron,i| neuron.output = self.vector[i] }
|
22
|
+
|
23
|
+
self.last_error = [1]
|
24
|
+
self.runs = 0
|
25
|
+
end
|
26
|
+
|
27
|
+
def associated?
|
28
|
+
return self.last_error.include? 0
|
29
|
+
end
|
30
|
+
|
31
|
+
def pattern
|
32
|
+
return self.state
|
33
|
+
end
|
34
|
+
|
35
|
+
def propagate
|
36
|
+
# Select random neuron
|
37
|
+
i = rand(self.neurons.size)
|
38
|
+
activation = 0
|
39
|
+
self.neurons.each_with_index do |other, j|
|
40
|
+
activation += other.weights[i]*other.output if i!=j
|
41
|
+
end
|
42
|
+
output = transfer(activation)
|
43
|
+
change = output != self.neurons[i].output
|
44
|
+
self.neurons[i].output = output
|
45
|
+
|
46
|
+
# Compile state of outputs
|
47
|
+
state = Array.new(self.neurons.size){|i| self.neurons[i].output}
|
48
|
+
|
49
|
+
# Calculate the current error
|
50
|
+
self.last_error = calculate_error(state)
|
51
|
+
|
52
|
+
# Convert state to binary and back to a multi dimensional array
|
53
|
+
state = to_binary(state)
|
54
|
+
state = state.each_slice(self.pattern_width).to_a
|
55
|
+
self.state = state
|
56
|
+
|
57
|
+
self.runs += 1
|
58
|
+
|
59
|
+
return {
|
60
|
+
:did_change => change,
|
61
|
+
:state => self.state,
|
62
|
+
:error => self.last_error
|
63
|
+
}
|
64
|
+
end
|
65
|
+
|
66
|
+
def calculate_error(current_pattern)
|
67
|
+
errors = Array.new(0)
|
68
|
+
|
69
|
+
self.patterns.each do |pattern|
|
70
|
+
sum = 0
|
71
|
+
|
72
|
+
expected = pattern.flatten
|
73
|
+
actual = current_pattern
|
74
|
+
|
75
|
+
expected.each_with_index do |v, i|
|
76
|
+
sum += 1 if expected[i]!=actual[i]
|
77
|
+
end
|
78
|
+
errors << sum
|
79
|
+
end
|
80
|
+
return errors
|
81
|
+
end
|
82
|
+
|
83
|
+
def transfer(activation)
|
84
|
+
(activation >= 0 ? 1 : -1)
|
85
|
+
end
|
86
|
+
|
87
|
+
def to_binary(vector)
|
88
|
+
return Array.new(vector.size){|i| ((vector[i]==-1) ? 0 : 1)}
|
89
|
+
end
|
90
|
+
|
91
|
+
end
|
92
|
+
end
|
@@ -0,0 +1,18 @@
|
|
1
|
+
module Hopfield
|
2
|
+
class Neuron
|
3
|
+
attr_accessor :weights, :output
|
4
|
+
|
5
|
+
def initialize(pattern_size)
|
6
|
+
minmax = Array.new(pattern_size) { [-0.5, 0.5] }
|
7
|
+
|
8
|
+
self.weights = random_vector(minmax)
|
9
|
+
end
|
10
|
+
|
11
|
+
def random_vector(minmax)
|
12
|
+
return Array.new(minmax.size) do |i|
|
13
|
+
minmax[i][0] + ((minmax[i][1] - minmax[i][0]) * rand())
|
14
|
+
end
|
15
|
+
end
|
16
|
+
|
17
|
+
end
|
18
|
+
end
|
@@ -0,0 +1,47 @@
|
|
1
|
+
module Hopfield
|
2
|
+
class Training
|
3
|
+
attr_accessor :patterns, :neurons, :pattern_width
|
4
|
+
|
5
|
+
def initialize(patterns)
|
6
|
+
# Check if patterns are the same size
|
7
|
+
unless patterns.map(&:size).uniq.count == 1
|
8
|
+
raise SyntaxError, 'Inconsistent pattern size'
|
9
|
+
end
|
10
|
+
|
11
|
+
# Turn 0 into -1
|
12
|
+
patterns.map { |pattern| pattern.map {|value| (value == 0 ? -1 : value) }}
|
13
|
+
|
14
|
+
# Set the patterns for this training
|
15
|
+
self.patterns = patterns
|
16
|
+
|
17
|
+
# Calculate the amount of required neurons
|
18
|
+
# This number is based on the number of inputs of a pattern
|
19
|
+
connections = patterns.first.map(&:size).inject{|sum,x| sum + x }
|
20
|
+
self.pattern_width = patterns.first.first.size
|
21
|
+
|
22
|
+
# Create neurons
|
23
|
+
self.neurons = Array.new(connections) { Neuron.new connections }
|
24
|
+
|
25
|
+
# Train the neurons
|
26
|
+
train(connections)
|
27
|
+
end
|
28
|
+
|
29
|
+
def train(connections)
|
30
|
+
self.neurons.each_with_index do |neuron, i|
|
31
|
+
for j in ((i+1)...connections) do
|
32
|
+
wij = 0.0
|
33
|
+
self.patterns.each do |pattern|
|
34
|
+
vector = pattern.flatten
|
35
|
+
#puts "Pattern: " + pattern.size.to_s
|
36
|
+
#puts "Pattern Y: " + pattern.first.size.to_s
|
37
|
+
#puts "Vector: " + vector.size.to_s
|
38
|
+
wij += vector[i]*vector[j]
|
39
|
+
end
|
40
|
+
self.neurons[i].weights[j] = wij
|
41
|
+
self.neurons[j].weights[i] = wij
|
42
|
+
end
|
43
|
+
end
|
44
|
+
end
|
45
|
+
|
46
|
+
end
|
47
|
+
end
|
metadata
ADDED
@@ -0,0 +1,85 @@
|
|
1
|
+
--- !ruby/object:Gem::Specification
|
2
|
+
name: hopfield
|
3
|
+
version: !ruby/object:Gem::Version
|
4
|
+
version: '1.0'
|
5
|
+
prerelease:
|
6
|
+
platform: ruby
|
7
|
+
authors:
|
8
|
+
- Bart Olsthoorn
|
9
|
+
autorequire:
|
10
|
+
bindir: bin
|
11
|
+
cert_chain: []
|
12
|
+
date: 2013-02-17 00:00:00.000000000 Z
|
13
|
+
dependencies:
|
14
|
+
- !ruby/object:Gem::Dependency
|
15
|
+
name: rspec
|
16
|
+
requirement: !ruby/object:Gem::Requirement
|
17
|
+
none: false
|
18
|
+
requirements:
|
19
|
+
- - ! '>='
|
20
|
+
- !ruby/object:Gem::Version
|
21
|
+
version: '0'
|
22
|
+
type: :development
|
23
|
+
prerelease: false
|
24
|
+
version_requirements: !ruby/object:Gem::Requirement
|
25
|
+
none: false
|
26
|
+
requirements:
|
27
|
+
- - ! '>='
|
28
|
+
- !ruby/object:Gem::Version
|
29
|
+
version: '0'
|
30
|
+
- !ruby/object:Gem::Dependency
|
31
|
+
name: chunky_png
|
32
|
+
requirement: !ruby/object:Gem::Requirement
|
33
|
+
none: false
|
34
|
+
requirements:
|
35
|
+
- - ! '>='
|
36
|
+
- !ruby/object:Gem::Version
|
37
|
+
version: '0'
|
38
|
+
type: :development
|
39
|
+
prerelease: false
|
40
|
+
version_requirements: !ruby/object:Gem::Requirement
|
41
|
+
none: false
|
42
|
+
requirements:
|
43
|
+
- - ! '>='
|
44
|
+
- !ruby/object:Gem::Version
|
45
|
+
version: '0'
|
46
|
+
description: Hopfield networks can be used for smart pattern recollections. It's recalling
|
47
|
+
patterns by modelling associative memory of a neural network
|
48
|
+
email:
|
49
|
+
- bartolsthoorn@gmail.com
|
50
|
+
executables: []
|
51
|
+
extensions: []
|
52
|
+
extra_rdoc_files: []
|
53
|
+
files:
|
54
|
+
- lib/hopfield/network.rb
|
55
|
+
- lib/hopfield/neuron.rb
|
56
|
+
- lib/hopfield/training.rb
|
57
|
+
- lib/hopfield.rb
|
58
|
+
- LICENSE
|
59
|
+
- README.md
|
60
|
+
homepage: http://github.com/bartolsthoorn/ruby-hopfield
|
61
|
+
licenses: []
|
62
|
+
post_install_message:
|
63
|
+
rdoc_options: []
|
64
|
+
require_paths:
|
65
|
+
- lib
|
66
|
+
required_ruby_version: !ruby/object:Gem::Requirement
|
67
|
+
none: false
|
68
|
+
requirements:
|
69
|
+
- - ! '>='
|
70
|
+
- !ruby/object:Gem::Version
|
71
|
+
version: '0'
|
72
|
+
required_rubygems_version: !ruby/object:Gem::Requirement
|
73
|
+
none: false
|
74
|
+
requirements:
|
75
|
+
- - ! '>='
|
76
|
+
- !ruby/object:Gem::Version
|
77
|
+
version: '0'
|
78
|
+
requirements: []
|
79
|
+
rubyforge_project:
|
80
|
+
rubygems_version: 1.8.24
|
81
|
+
signing_key:
|
82
|
+
specification_version: 3
|
83
|
+
summary: Ruby implementation of a Hopfield Network
|
84
|
+
test_files: []
|
85
|
+
has_rdoc:
|