hmm 0.0.1
Sign up to get free protection for your applications and to get access to all the features.
- data/.document +5 -0
- data/.gitignore +21 -0
- data/LICENSE +20 -0
- data/README.rdoc +17 -0
- data/Rakefile +54 -0
- data/VERSION +1 -0
- data/hmm.gemspec +57 -0
- data/lib/hmm.rb +135 -0
- data/test/helper.rb +10 -0
- data/test/test_hmm.rb +7 -0
- metadata +85 -0
data/.document
ADDED
data/.gitignore
ADDED
data/LICENSE
ADDED
@@ -0,0 +1,20 @@
|
|
1
|
+
Copyright (c) 2009 David Tresner-Kirsch
|
2
|
+
|
3
|
+
Permission is hereby granted, free of charge, to any person obtaining
|
4
|
+
a copy of this software and associated documentation files (the
|
5
|
+
"Software"), to deal in the Software without restriction, including
|
6
|
+
without limitation the rights to use, copy, modify, merge, publish,
|
7
|
+
distribute, sublicense, and/or sell copies of the Software, and to
|
8
|
+
permit persons to whom the Software is furnished to do so, subject to
|
9
|
+
the following conditions:
|
10
|
+
|
11
|
+
The above copyright notice and this permission notice shall be
|
12
|
+
included in all copies or substantial portions of the Software.
|
13
|
+
|
14
|
+
THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
|
15
|
+
EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
|
16
|
+
MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND
|
17
|
+
NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS BE
|
18
|
+
LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION
|
19
|
+
OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION
|
20
|
+
WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
|
data/README.rdoc
ADDED
@@ -0,0 +1,17 @@
|
|
1
|
+
= hmm
|
2
|
+
|
3
|
+
Description goes here.
|
4
|
+
|
5
|
+
== Note on Patches/Pull Requests
|
6
|
+
|
7
|
+
* Fork the project.
|
8
|
+
* Make your feature addition or bug fix.
|
9
|
+
* Add tests for it. This is important so I don't break it in a
|
10
|
+
future version unintentionally.
|
11
|
+
* Commit, do not mess with rakefile, version, or history.
|
12
|
+
(if you want to have your own version, that is fine but bump version in a commit by itself I can ignore when I pull)
|
13
|
+
* Send me a pull request. Bonus points for topic branches.
|
14
|
+
|
15
|
+
== Copyright
|
16
|
+
|
17
|
+
Copyright (c) 2009 David Tresner-Kirsch. See LICENSE for details.
|
data/Rakefile
ADDED
@@ -0,0 +1,54 @@
|
|
1
|
+
require 'rubygems'
|
2
|
+
require 'rake'
|
3
|
+
|
4
|
+
begin
|
5
|
+
require 'jeweler'
|
6
|
+
Jeweler::Tasks.new do |gem|
|
7
|
+
gem.name = "hmm"
|
8
|
+
gem.summary = %Q{HMM Classifier}
|
9
|
+
gem.description = %Q{This project is a Ruby gem ('hmm') for machine learning that natively implements a (somewhat) generalized Hidden Markov Model classifier.}
|
10
|
+
gem.email = "dwkirsch@gmail.com"
|
11
|
+
gem.homepage = "http://github.com/dtkirsch/hmm"
|
12
|
+
gem.authors = ["David Tresner-Kirsch"]
|
13
|
+
gem.add_development_dependency "thoughtbot-shoulda", ">= 0"
|
14
|
+
gem.add_development_dependency "narray", ">= 0"
|
15
|
+
# gem is a Gem::Specification... see http://www.rubygems.org/read/chapter/20 for additional settings
|
16
|
+
end
|
17
|
+
Jeweler::GemcutterTasks.new
|
18
|
+
rescue LoadError
|
19
|
+
puts "Jeweler (or a dependency) not available. Install it with: gem install jeweler"
|
20
|
+
end
|
21
|
+
|
22
|
+
require 'rake/testtask'
|
23
|
+
Rake::TestTask.new(:test) do |test|
|
24
|
+
test.libs << 'lib' << 'test'
|
25
|
+
test.pattern = 'test/**/test_*.rb'
|
26
|
+
test.verbose = true
|
27
|
+
end
|
28
|
+
|
29
|
+
begin
|
30
|
+
require 'rcov/rcovtask'
|
31
|
+
Rcov::RcovTask.new do |test|
|
32
|
+
test.libs << 'test'
|
33
|
+
test.pattern = 'test/**/test_*.rb'
|
34
|
+
test.verbose = true
|
35
|
+
end
|
36
|
+
rescue LoadError
|
37
|
+
task :rcov do
|
38
|
+
abort "RCov is not available. In order to run rcov, you must: sudo gem install spicycode-rcov"
|
39
|
+
end
|
40
|
+
end
|
41
|
+
|
42
|
+
task :test => :check_dependencies
|
43
|
+
|
44
|
+
task :default => :test
|
45
|
+
|
46
|
+
require 'rake/rdoctask'
|
47
|
+
Rake::RDocTask.new do |rdoc|
|
48
|
+
version = File.exist?('VERSION') ? File.read('VERSION') : ""
|
49
|
+
|
50
|
+
rdoc.rdoc_dir = 'rdoc'
|
51
|
+
rdoc.title = "hmm #{version}"
|
52
|
+
rdoc.rdoc_files.include('README*')
|
53
|
+
rdoc.rdoc_files.include('lib/**/*.rb')
|
54
|
+
end
|
data/VERSION
ADDED
@@ -0,0 +1 @@
|
|
1
|
+
0.0.1
|
data/hmm.gemspec
ADDED
@@ -0,0 +1,57 @@
|
|
1
|
+
# Generated by jeweler
|
2
|
+
# DO NOT EDIT THIS FILE DIRECTLY
|
3
|
+
# Instead, edit Jeweler::Tasks in Rakefile, and run the gemspec command
|
4
|
+
# -*- encoding: utf-8 -*-
|
5
|
+
|
6
|
+
Gem::Specification.new do |s|
|
7
|
+
s.name = %q{hmm}
|
8
|
+
s.version = "0.0.1"
|
9
|
+
|
10
|
+
s.required_rubygems_version = Gem::Requirement.new(">= 0") if s.respond_to? :required_rubygems_version=
|
11
|
+
s.authors = ["David Tresner-Kirsch"]
|
12
|
+
s.date = %q{2009-11-25}
|
13
|
+
s.description = %q{This project is a Ruby gem ('hmm') for machine learning that natively implements a (somewhat) generalized Hidden Markov Model classifier.}
|
14
|
+
s.email = %q{dwkirsch@gmail.com}
|
15
|
+
s.extra_rdoc_files = [
|
16
|
+
"LICENSE",
|
17
|
+
"README.rdoc"
|
18
|
+
]
|
19
|
+
s.files = [
|
20
|
+
".document",
|
21
|
+
".gitignore",
|
22
|
+
"LICENSE",
|
23
|
+
"README.rdoc",
|
24
|
+
"Rakefile",
|
25
|
+
"VERSION",
|
26
|
+
"hmm.gemspec",
|
27
|
+
"lib/hmm.rb",
|
28
|
+
"test/helper.rb",
|
29
|
+
"test/test_hmm.rb"
|
30
|
+
]
|
31
|
+
s.homepage = %q{http://github.com/dtkirsch/hmm}
|
32
|
+
s.rdoc_options = ["--charset=UTF-8"]
|
33
|
+
s.require_paths = ["lib"]
|
34
|
+
s.rubygems_version = %q{1.3.5}
|
35
|
+
s.summary = %q{HMM Classifier}
|
36
|
+
s.test_files = [
|
37
|
+
"test/helper.rb",
|
38
|
+
"test/test_hmm.rb"
|
39
|
+
]
|
40
|
+
|
41
|
+
if s.respond_to? :specification_version then
|
42
|
+
current_version = Gem::Specification::CURRENT_SPECIFICATION_VERSION
|
43
|
+
s.specification_version = 3
|
44
|
+
|
45
|
+
if Gem::Version.new(Gem::RubyGemsVersion) >= Gem::Version.new('1.2.0') then
|
46
|
+
s.add_development_dependency(%q<thoughtbot-shoulda>, [">= 0"])
|
47
|
+
s.add_development_dependency(%q<narray>, [">= 0"])
|
48
|
+
else
|
49
|
+
s.add_dependency(%q<thoughtbot-shoulda>, [">= 0"])
|
50
|
+
s.add_dependency(%q<narray>, [">= 0"])
|
51
|
+
end
|
52
|
+
else
|
53
|
+
s.add_dependency(%q<thoughtbot-shoulda>, [">= 0"])
|
54
|
+
s.add_dependency(%q<narray>, [">= 0"])
|
55
|
+
end
|
56
|
+
end
|
57
|
+
|
data/lib/hmm.rb
ADDED
@@ -0,0 +1,135 @@
|
|
1
|
+
# Hidden Markov Model classifier.
|
2
|
+
# Currently capable of:
|
3
|
+
# -supervised training on data with arbitrary state and observation domains.
|
4
|
+
# -decoding of obsservation strings via Viterbi
|
5
|
+
# -computing token level accuracy across a list of observation sequences
|
6
|
+
# against a provided gold standard
|
7
|
+
|
8
|
+
|
9
|
+
require 'rubygems'
|
10
|
+
require 'narray'
|
11
|
+
|
12
|
+
class HMM
|
13
|
+
|
14
|
+
class Classifier
|
15
|
+
attr_accessor :a, :b, :pi, :o_lex, :q_lex, :debug, :train
|
16
|
+
# Member variables:
|
17
|
+
# pi -- initial state distribution
|
18
|
+
# a -- state transition probabilities
|
19
|
+
# b -- state-conditional observation probabilities
|
20
|
+
# o_lex -- index of observation labels
|
21
|
+
# q_lex -- index of state labels
|
22
|
+
# debug -- flag for verbose output to stdout
|
23
|
+
# train -- a list of labelled sequences for supervised training
|
24
|
+
|
25
|
+
def initialize
|
26
|
+
@o_lex, @q_lex, @train = [], [], []
|
27
|
+
end
|
28
|
+
|
29
|
+
def add_to_train(o, q)
|
30
|
+
@o_lex |= o # add new tokens to indexed lexicon
|
31
|
+
@q_lex |= q
|
32
|
+
@train << Sequence.new(index(o, @o_lex), index(q, @q_lex))
|
33
|
+
end
|
34
|
+
|
35
|
+
def train
|
36
|
+
# initialize Pi, A, and B
|
37
|
+
@pi = NArray.float(@q_lex.length)
|
38
|
+
@a = NArray.float(@q_lex.length, @q_lex.length)
|
39
|
+
@b = NArray.float(@q_lex.length, @o_lex.length)
|
40
|
+
|
41
|
+
# count frequencies to build Pi, A, and B
|
42
|
+
@train.each do |sequence|
|
43
|
+
@pi[sequence.q.first] +=1
|
44
|
+
sequence.q.length.times do |i|
|
45
|
+
@b[sequence.q[i], sequence.o[i]] += 1
|
46
|
+
@a[sequence.q[i-1], sequence.q[i]] +=1 if i>0
|
47
|
+
end
|
48
|
+
end
|
49
|
+
|
50
|
+
# normalize frequencies into probabilities
|
51
|
+
@pi /= @pi.sum
|
52
|
+
@a /= @a.sum(1)
|
53
|
+
@b /= @b.sum(1)
|
54
|
+
end
|
55
|
+
|
56
|
+
|
57
|
+
def decode(o_sequence)
|
58
|
+
# Viterbi! with log probability math to avoid underflow
|
59
|
+
|
60
|
+
# encode observations
|
61
|
+
o_sequence = index(o_sequence, @o_lex)
|
62
|
+
|
63
|
+
# initialize. skipping the 0 initialization for psi, as it's never used.
|
64
|
+
# psi will have T-1 elements instead of T, allowing it
|
65
|
+
# to control the backtrack iterator later.
|
66
|
+
delta, psi = [log(pi)+log(b[true, o_sequence.shift])], []
|
67
|
+
|
68
|
+
# recursive step
|
69
|
+
o_sequence.each do |o|
|
70
|
+
psi << argmax(delta.last+log(a))
|
71
|
+
delta << (delta.last+log(a)).max(0)+log(b[true, o])
|
72
|
+
end
|
73
|
+
|
74
|
+
# initialize Q* with final state
|
75
|
+
q_star = [delta.last.sort_index[-1]]
|
76
|
+
|
77
|
+
# backtrack the optimal state sequence into Q*
|
78
|
+
psi.reverse.each do |psi_t|
|
79
|
+
q_star.unshift psi_t[q_star.first]
|
80
|
+
end
|
81
|
+
|
82
|
+
puts "delta:", exp(delta).inspect, "psi:", exp(psi).inspect if @debug
|
83
|
+
|
84
|
+
return deindex(q_star, @q_lex)
|
85
|
+
end
|
86
|
+
|
87
|
+
def accuracy(o, q)
|
88
|
+
# token level accuracy across a set of sequences
|
89
|
+
correct, total = 0.0, 0.0
|
90
|
+
o.length.times do |i|
|
91
|
+
correct += (NArray.to_na(decode(o[i])).eq NArray.to_na(q[i])).sum
|
92
|
+
total += o[i].length
|
93
|
+
end
|
94
|
+
correct/total
|
95
|
+
end
|
96
|
+
|
97
|
+
private
|
98
|
+
|
99
|
+
# index and deindex map between labels and the ordinals of those labels.
|
100
|
+
# the ordinals map the labels to rows and columns of Pi, A, and B
|
101
|
+
def index(sequence, lexicon)
|
102
|
+
lexicon |= sequence # add any unknown tokens to the lex
|
103
|
+
sequence.collect{|x| lexicon.rindex(x)}
|
104
|
+
end
|
105
|
+
|
106
|
+
def deindex(sequence, lexicon)
|
107
|
+
sequence.collect{|i| lexicon[i]}
|
108
|
+
end
|
109
|
+
|
110
|
+
# abstracting out some array element operations for readability
|
111
|
+
def log(array)
|
112
|
+
# natural log of each element
|
113
|
+
array.collect{|n| NMath::log n}
|
114
|
+
end
|
115
|
+
|
116
|
+
def exp(array)
|
117
|
+
# e to the power of each element
|
118
|
+
array.collect{|n| Math::E ** n}
|
119
|
+
end
|
120
|
+
|
121
|
+
def argmax(narray)
|
122
|
+
# horizontal index of the max in each row.
|
123
|
+
# the mod is b/c sort_index returns global indices
|
124
|
+
# (rather than starting at 0 for each row)
|
125
|
+
(narray).sort_index(0)[-1, true] % narray.shape[1]
|
126
|
+
end
|
127
|
+
end
|
128
|
+
|
129
|
+
class Sequence
|
130
|
+
attr_accessor :o, :q # array of observations, array of states
|
131
|
+
def initialize (o, q)
|
132
|
+
@o, @q = o, q
|
133
|
+
end
|
134
|
+
end
|
135
|
+
end
|
data/test/helper.rb
ADDED
data/test/test_hmm.rb
ADDED
metadata
ADDED
@@ -0,0 +1,85 @@
|
|
1
|
+
--- !ruby/object:Gem::Specification
|
2
|
+
name: hmm
|
3
|
+
version: !ruby/object:Gem::Version
|
4
|
+
version: 0.0.1
|
5
|
+
platform: ruby
|
6
|
+
authors:
|
7
|
+
- David Tresner-Kirsch
|
8
|
+
autorequire:
|
9
|
+
bindir: bin
|
10
|
+
cert_chain: []
|
11
|
+
|
12
|
+
date: 2009-11-25 00:00:00 -05:00
|
13
|
+
default_executable:
|
14
|
+
dependencies:
|
15
|
+
- !ruby/object:Gem::Dependency
|
16
|
+
name: thoughtbot-shoulda
|
17
|
+
type: :development
|
18
|
+
version_requirement:
|
19
|
+
version_requirements: !ruby/object:Gem::Requirement
|
20
|
+
requirements:
|
21
|
+
- - ">="
|
22
|
+
- !ruby/object:Gem::Version
|
23
|
+
version: "0"
|
24
|
+
version:
|
25
|
+
- !ruby/object:Gem::Dependency
|
26
|
+
name: narray
|
27
|
+
type: :development
|
28
|
+
version_requirement:
|
29
|
+
version_requirements: !ruby/object:Gem::Requirement
|
30
|
+
requirements:
|
31
|
+
- - ">="
|
32
|
+
- !ruby/object:Gem::Version
|
33
|
+
version: "0"
|
34
|
+
version:
|
35
|
+
description: This project is a Ruby gem ('hmm') for machine learning that natively implements a (somewhat) generalized Hidden Markov Model classifier.
|
36
|
+
email: dwkirsch@gmail.com
|
37
|
+
executables: []
|
38
|
+
|
39
|
+
extensions: []
|
40
|
+
|
41
|
+
extra_rdoc_files:
|
42
|
+
- LICENSE
|
43
|
+
- README.rdoc
|
44
|
+
files:
|
45
|
+
- .document
|
46
|
+
- .gitignore
|
47
|
+
- LICENSE
|
48
|
+
- README.rdoc
|
49
|
+
- Rakefile
|
50
|
+
- VERSION
|
51
|
+
- hmm.gemspec
|
52
|
+
- lib/hmm.rb
|
53
|
+
- test/helper.rb
|
54
|
+
- test/test_hmm.rb
|
55
|
+
has_rdoc: true
|
56
|
+
homepage: http://github.com/dtkirsch/hmm
|
57
|
+
licenses: []
|
58
|
+
|
59
|
+
post_install_message:
|
60
|
+
rdoc_options:
|
61
|
+
- --charset=UTF-8
|
62
|
+
require_paths:
|
63
|
+
- lib
|
64
|
+
required_ruby_version: !ruby/object:Gem::Requirement
|
65
|
+
requirements:
|
66
|
+
- - ">="
|
67
|
+
- !ruby/object:Gem::Version
|
68
|
+
version: "0"
|
69
|
+
version:
|
70
|
+
required_rubygems_version: !ruby/object:Gem::Requirement
|
71
|
+
requirements:
|
72
|
+
- - ">="
|
73
|
+
- !ruby/object:Gem::Version
|
74
|
+
version: "0"
|
75
|
+
version:
|
76
|
+
requirements: []
|
77
|
+
|
78
|
+
rubyforge_project:
|
79
|
+
rubygems_version: 1.3.5
|
80
|
+
signing_key:
|
81
|
+
specification_version: 3
|
82
|
+
summary: HMM Classifier
|
83
|
+
test_files:
|
84
|
+
- test/helper.rb
|
85
|
+
- test/test_hmm.rb
|