hmm 0.0.1

data/.document

@@ -0,0 +1,5 @@
+README.rdoc
+lib/**/*.rb
+bin/*
+features/**/*.feature
+LICENSE

data/.gitignore

@@ -0,0 +1,21 @@
+## MAC OS
+.DS_Store
+
+## TEXTMATE
+*.tmproj
+tmtags
+
+## EMACS
+*~
+\#*
+.\#*
+
+## VIM
+*.swp
+
+## PROJECT::GENERAL
+coverage
+rdoc
+pkg
+
+## PROJECT::SPECIFIC

data/LICENSE

@@ -0,0 +1,20 @@
+Copyright (c) 2009 David Tresner-Kirsch
+
+Permission is hereby granted, free of charge, to any person obtaining
+a copy of this software and associated documentation files (the
+"Software"), to deal in the Software without restriction, including
+without limitation the rights to use, copy, modify, merge, publish,
+distribute, sublicense, and/or sell copies of the Software, and to
+permit persons to whom the Software is furnished to do so, subject to
+the following conditions:
+
+The above copyright notice and this permission notice shall be
+included in all copies or substantial portions of the Software.
+
+THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
+EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND
+NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS BE
+LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION
+OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION
+WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.

data/README.rdoc

@@ -0,0 +1,17 @@
+= hmm
+
+Description goes here.
+
+== Note on Patches/Pull Requests
+ 
+* Fork the project.
+* Make your feature addition or bug fix.
+* Add tests for it. This is important so I don't break it in a
+  future version unintentionally.
+* Commit, do not mess with rakefile, version, or history.
+  (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)
+* Send me a pull request. Bonus points for topic branches.
+
+== Copyright
+
+Copyright (c) 2009 David Tresner-Kirsch. See LICENSE for details.

data/Rakefile

@@ -0,0 +1,54 @@
+require 'rubygems'
+require 'rake'
+
+begin
+  require 'jeweler'
+  Jeweler::Tasks.new do |gem|
+    gem.name = "hmm"
+    gem.summary = %Q{HMM Classifier}
+    gem.description = %Q{This project is a Ruby gem ('hmm') for machine learning that natively implements a (somewhat) generalized Hidden Markov Model classifier.}
+    gem.email = "dwkirsch@gmail.com"
+    gem.homepage = "http://github.com/dtkirsch/hmm"
+    gem.authors = ["David Tresner-Kirsch"]
+    gem.add_development_dependency "thoughtbot-shoulda", ">= 0"
+    gem.add_development_dependency "narray", ">= 0"
+    # gem is a Gem::Specification... see http://www.rubygems.org/read/chapter/20 for additional settings
+  end
+  Jeweler::GemcutterTasks.new
+rescue LoadError
+  puts "Jeweler (or a dependency) not available. Install it with: gem install jeweler"
+end
+
+require 'rake/testtask'
+Rake::TestTask.new(:test) do |test|
+  test.libs << 'lib' << 'test'
+  test.pattern = 'test/**/test_*.rb'
+  test.verbose = true
+end
+
+begin
+  require 'rcov/rcovtask'
+  Rcov::RcovTask.new do |test|
+    test.libs << 'test'
+    test.pattern = 'test/**/test_*.rb'
+    test.verbose = true
+  end
+rescue LoadError
+  task :rcov do
+    abort "RCov is not available. In order to run rcov, you must: sudo gem install spicycode-rcov"
+  end
+end
+
+task :test => :check_dependencies
+
+task :default => :test
+
+require 'rake/rdoctask'
+Rake::RDocTask.new do |rdoc|
+  version = File.exist?('VERSION') ? File.read('VERSION') : ""
+
+  rdoc.rdoc_dir = 'rdoc'
+  rdoc.title = "hmm #{version}"
+  rdoc.rdoc_files.include('README*')
+  rdoc.rdoc_files.include('lib/**/*.rb')
+end

data/VERSION

@@ -0,0 +1 @@
+0.0.1

data/hmm.gemspec

@@ -0,0 +1,57 @@
+# Generated by jeweler
+# DO NOT EDIT THIS FILE DIRECTLY
+# Instead, edit Jeweler::Tasks in Rakefile, and run the gemspec command
+# -*- encoding: utf-8 -*-
+
+Gem::Specification.new do |s|
+  s.name = %q{hmm}
+  s.version = "0.0.1"
+
+  s.required_rubygems_version = Gem::Requirement.new(">= 0") if s.respond_to? :required_rubygems_version=
+  s.authors = ["David Tresner-Kirsch"]
+  s.date = %q{2009-11-25}
+  s.description = %q{This project is a Ruby gem ('hmm') for machine learning that natively implements a (somewhat) generalized Hidden Markov Model classifier.}
+  s.email = %q{dwkirsch@gmail.com}
+  s.extra_rdoc_files = [
+    "LICENSE",
+     "README.rdoc"
+  ]
+  s.files = [
+    ".document",
+     ".gitignore",
+     "LICENSE",
+     "README.rdoc",
+     "Rakefile",
+     "VERSION",
+     "hmm.gemspec",
+     "lib/hmm.rb",
+     "test/helper.rb",
+     "test/test_hmm.rb"
+  ]
+  s.homepage = %q{http://github.com/dtkirsch/hmm}
+  s.rdoc_options = ["--charset=UTF-8"]
+  s.require_paths = ["lib"]
+  s.rubygems_version = %q{1.3.5}
+  s.summary = %q{HMM Classifier}
+  s.test_files = [
+    "test/helper.rb",
+     "test/test_hmm.rb"
+  ]
+
+  if s.respond_to? :specification_version then
+    current_version = Gem::Specification::CURRENT_SPECIFICATION_VERSION
+    s.specification_version = 3
+
+    if Gem::Version.new(Gem::RubyGemsVersion) >= Gem::Version.new('1.2.0') then
+      s.add_development_dependency(%q<thoughtbot-shoulda>, [">= 0"])
+      s.add_development_dependency(%q<narray>, [">= 0"])
+    else
+      s.add_dependency(%q<thoughtbot-shoulda>, [">= 0"])
+      s.add_dependency(%q<narray>, [">= 0"])
+    end
+  else
+    s.add_dependency(%q<thoughtbot-shoulda>, [">= 0"])
+    s.add_dependency(%q<narray>, [">= 0"])
+  end
+end
+

data/lib/hmm.rb

@@ -0,0 +1,135 @@
+# Hidden Markov Model classifier.
+# Currently capable of:
+#	-supervised training on data with arbitrary state and observation domains.
+# 	-decoding of obsservation strings via Viterbi
+#	-computing token level accuracy across a list of observation sequences
+#		against a provided gold standard
+
+
+require 'rubygems'
+require 'narray'
+
+class HMM
+
+	class Classifier
+		attr_accessor :a, :b, :pi, :o_lex, :q_lex, :debug, :train
+		# Member variables:
+		# 	pi -- initial state distribution
+		#	a -- state transition probabilities
+		#	b -- state-conditional observation probabilities
+		#	o_lex -- index of observation labels
+		#	q_lex -- index of state labels
+		#	debug -- flag for verbose output to stdout
+		#	train -- a list of labelled sequences for supervised training
+		
+		def initialize
+			@o_lex, @q_lex, @train = [], [], []
+		end
+		
+		def add_to_train(o, q)
+			@o_lex |= o # add new tokens to indexed lexicon
+			@q_lex |= q
+			@train << Sequence.new(index(o, @o_lex), index(q, @q_lex))
+		end
+		
+		def train
+			# initialize Pi, A, and B
+			@pi = NArray.float(@q_lex.length)
+			@a = NArray.float(@q_lex.length, @q_lex.length)
+			@b = NArray.float(@q_lex.length, @o_lex.length)
+			
+			# count frequencies to build Pi, A, and B
+			@train.each do |sequence|
+				@pi[sequence.q.first] +=1
+				sequence.q.length.times do |i|
+					@b[sequence.q[i], sequence.o[i]] += 1
+					@a[sequence.q[i-1], sequence.q[i]] +=1 if i>0
+				end
+			end
+			
+			# normalize frequencies into probabilities
+			@pi /= @pi.sum
+			@a /= @a.sum(1)
+			@b /= @b.sum(1)
+		end
+	    
+
+		def decode(o_sequence)
+			# Viterbi!  with log probability math to avoid underflow
+			
+			# encode observations
+			o_sequence = index(o_sequence, @o_lex)
+			
+			# initialize.  skipping the 0 initialization for psi, as it's never used.
+			# psi will have T-1 elements instead of T, allowing it
+			# to control the backtrack iterator later.
+			delta, psi = [log(pi)+log(b[true, o_sequence.shift])], []
+
+			# recursive step
+			o_sequence.each do |o|
+				psi << argmax(delta.last+log(a))
+				delta << (delta.last+log(a)).max(0)+log(b[true, o])
+			end
+			
+			# initialize Q* with final state
+			q_star = [delta.last.sort_index[-1]]
+			
+			# backtrack the optimal state sequence into Q*
+			psi.reverse.each do |psi_t|
+				q_star.unshift psi_t[q_star.first]
+			end
+			
+			puts "delta:", exp(delta).inspect, "psi:", exp(psi).inspect if @debug
+			
+			return deindex(q_star, @q_lex)
+		end
+		
+		def accuracy(o, q)
+			# token level accuracy across a set of sequences
+			correct, total = 0.0, 0.0
+			o.length.times do |i|
+				correct += (NArray.to_na(decode(o[i])).eq NArray.to_na(q[i])).sum
+				total += o[i].length
+			end
+			correct/total
+		end
+		
+		private
+		
+		# index and deindex map between labels and the ordinals of those labels.
+		# the ordinals map the labels to rows and columns of Pi, A, and B
+		def index(sequence, lexicon)
+                        lexicon |= sequence # add any unknown tokens to the lex
+			sequence.collect{|x| lexicon.rindex(x)}
+		end
+		
+		def deindex(sequence, lexicon)
+			sequence.collect{|i| lexicon[i]}
+		end
+		
+		# abstracting out some array element operations for readability
+		def log(array)
+			# natural log of each element
+			array.collect{|n| NMath::log n}
+		end
+		
+		def exp(array)
+			# e to the power of each element
+			array.collect{|n| Math::E ** n}
+		end
+		
+		def argmax(narray)
+			# horizontal index of the max in each row.
+			# the mod is b/c sort_index returns global indices
+			# (rather than starting at 0 for each row)
+			(narray).sort_index(0)[-1, true] % narray.shape[1]
+		end
+	end
+
+	class Sequence
+	      attr_accessor :o, :q # array of observations, array of states
+	      def initialize (o, q)
+	      	  @o, @q = o, q
+	      end
+	end
+end

data/test/helper.rb

@@ -0,0 +1,10 @@
+require 'rubygems'
+require 'test/unit'
+require 'shoulda'
+
+$LOAD_PATH.unshift(File.join(File.dirname(__FILE__), '..', 'lib'))
+$LOAD_PATH.unshift(File.dirname(__FILE__))
+require 'hmm'
+
+class Test::Unit::TestCase
+end

data/test/test_hmm.rb

@@ -0,0 +1,7 @@
+require 'helper'
+
+class TestHmm < Test::Unit::TestCase
+  should "probably rename this file and start testing for real" do
+    #flunk "hey buddy, you should probably rename this file and start testing for real"
+  end
+end

metadata

@@ -0,0 +1,85 @@
+--- !ruby/object:Gem::Specification 
+name: hmm
+version: !ruby/object:Gem::Version 
+  version: 0.0.1
+platform: ruby
+authors: 
+- David Tresner-Kirsch
+autorequire: 
+bindir: bin
+cert_chain: []
+
+date: 2009-11-25 00:00:00 -05:00
+default_executable: 
+dependencies: 
+- !ruby/object:Gem::Dependency 
+  name: thoughtbot-shoulda
+  type: :development
+  version_requirement: 
+  version_requirements: !ruby/object:Gem::Requirement 
+    requirements: 
+    - - ">="
+      - !ruby/object:Gem::Version 
+        version: "0"
+    version: 
+- !ruby/object:Gem::Dependency 
+  name: narray
+  type: :development
+  version_requirement: 
+  version_requirements: !ruby/object:Gem::Requirement 
+    requirements: 
+    - - ">="
+      - !ruby/object:Gem::Version 
+        version: "0"
+    version: 
+description: This project is a Ruby gem ('hmm') for machine learning that natively implements a (somewhat) generalized Hidden Markov Model classifier.
+email: dwkirsch@gmail.com
+executables: []
+
+extensions: []
+
+extra_rdoc_files: 
+- LICENSE
+- README.rdoc
+files: 
+- .document
+- .gitignore
+- LICENSE
+- README.rdoc
+- Rakefile
+- VERSION
+- hmm.gemspec
+- lib/hmm.rb
+- test/helper.rb
+- test/test_hmm.rb
+has_rdoc: true
+homepage: http://github.com/dtkirsch/hmm
+licenses: []
+
+post_install_message: 
+rdoc_options: 
+- --charset=UTF-8
+require_paths: 
+- lib
+required_ruby_version: !ruby/object:Gem::Requirement 
+  requirements: 
+  - - ">="
+    - !ruby/object:Gem::Version 
+      version: "0"
+  version: 
+required_rubygems_version: !ruby/object:Gem::Requirement 
+  requirements: 
+  - - ">="
+    - !ruby/object:Gem::Version 
+      version: "0"
+  version: 
+requirements: []
+
+rubyforge_project: 
+rubygems_version: 1.3.5
+signing_key: 
+specification_version: 3
+summary: HMM Classifier
+test_files: 
+- test/helper.rb
+- test/test_hmm.rb