finite_mdp 0.3.0 → 0.4.0

checksums.yaml

@@ -1,7 +1,7 @@
 ---
 SHA1:
-  metadata.gz: abf81db7b691f5238c134d835f819f75609fa17c
-  data.tar.gz: 4d897e26e7cc8e8aaffd5c5ce80855d3de55fbb2
+  metadata.gz: bd00a14ccd84691b9ba6f544d1c73453e0cf6b68
+  data.tar.gz: 21a81b94680509a011ddf870bd6e7f86e905d000
 SHA512:
-  metadata.gz: 8711791575db42460dc233ab92a787697731e833c3f121a351ddb1a3dc690c411ec00ff99a0dd996b33fd1be6cae4ea1a423354c013e709f4cb12968b1e3d0c8
-  data.tar.gz: aae51101f51e9d60f9b648b58a3ea6443f8228b8ddee4b00c1d55861d8fd89f8c48d4b35174c4fbda2dbdff037d7631d64ecc3f45957e529e5720087e43fe2b3
+  metadata.gz: af3127d8c5d7d84260e143e18fd611914a767075c903f97b767dfb6b3654e0c1efd54e0c8cc147ab8c61b9a23c241a19abe0939255d520fb74bf0b1e46758019
+  data.tar.gz: 2f22ebc9a7bfacfb08c92c309e672da77a0dadb161be4c1a40e85f80fb690b67aad0839e27eb0ceb3b3404ea49eb16fb9344c82442ae700cd8a983201e8e37a1

data/lib/finite_mdp.rb

@@ -1,5 +1,4 @@
 # frozen_string_literal: true
-require 'enumerator'
 
 require 'finite_mdp/version'
 require 'finite_mdp/vector_valued'

data/lib/finite_mdp/array_model.rb

@@ -1,4 +1,5 @@
 # frozen_string_literal: true
+
 #
 # A finite markov decision process model for which the states, transition
 # probabilities and rewards are stored in a sparse nested array format:
@@ -216,7 +217,9 @@ class FiniteMDP::ArrayModel
           pr = model.transition_probability(state, action, next_state)
           next unless pr > 0 || !sparse
           reward = model.reward(state, action, next_state)
-          [state_action_map.state_index(next_state), pr, reward]
+          next_index = state_action_map.state_index(next_state)
+          raise "successor state not found: #{next_state}" unless next_index
+          [next_index, pr, reward]
         end.compact
       end
     end

data/lib/finite_mdp/hash_model.rb

@@ -1,4 +1,5 @@
 # frozen_string_literal: true
+
 #
 # A finite markov decision process model for which the transition
 # probabilities and rewards are specified using nested hash tables.

data/lib/finite_mdp/model.rb

@@ -1,4 +1,5 @@
 # frozen_string_literal: true
+
 #
 # Interface that defines a finite markov decision process model.
 #
@@ -202,8 +203,9 @@ module FiniteMDP::Model
   #
   def check_transition_probabilities_sum(tol = 1e-6)
     transition_probability_sums.each do |(state, action), pr|
-      raise "transition probabilities for state #{state.inspect} and
-          action #{action.inspect} sum to #{pr}" if pr < 1 - tol
+      next if (1 - pr).abs <= tol
+      raise "transition probabilities for state #{state.inspect} and action " \
+          "#{action.inspect} sum to #{pr}"
     end
     nil
   end

data/lib/finite_mdp/solver.rb

@@ -1,8 +1,8 @@
 # frozen_string_literal: true
 
 # We use A to denote a matrix, which rubocop does not like.
-# rubocop:disable Style/MethodName
-# rubocop:disable Style/VariableName
+# rubocop:disable Naming/MethodName
+# rubocop:disable Naming/VariableName
 
 require 'narray'
 
@@ -328,8 +328,10 @@ class FiniteMDP::Solver
       loop do
         value_delta = evaluate_policy
         num_value_iters += 1
-        yield(num_policy_iters, num_actions_changed, num_value_iters,
-          value_delta) if block_given?
+        if block_given?
+          yield(num_policy_iters, num_actions_changed, num_value_iters,
+            value_delta)
+        end
 
         break if value_delta < value_tolerance
         break if max_value_iters && num_value_iters >= max_value_iters
@@ -402,3 +404,6 @@ class FiniteMDP::Solver
     @policy_b[state_n] = b_n
   end
 end
+
+# rubocop:enable Naming/MethodName
+# rubocop:enable Naming/VariableName

data/lib/finite_mdp/table_model.rb

@@ -1,4 +1,5 @@
 # frozen_string_literal: true
+
 #
 # A finite markov decision process model for which the states, actions,
 # transition probabilities and rewards are specified as a table. This is a

data/lib/finite_mdp/vector_valued.rb

@@ -1,4 +1,5 @@
 # frozen_string_literal: true
+
 #
 # Define an object's hash code and equality (in the sense of <tt>eql?</tt>)
 # according to its array representation (<tt>to_a</tt>). See notes for {Model}

data/lib/finite_mdp/version.rb

@@ -1,7 +1,8 @@
 # frozen_string_literal: true
+
 module FiniteMDP
   VERSION_MAJOR = 0
-  VERSION_MINOR = 3
+  VERSION_MINOR = 4
   VERSION_PATCH = 0
   VERSION = [VERSION_MAJOR, VERSION_MINOR, VERSION_PATCH].join('.')
 end

data/test/finite_mdp/finite_mdp_test.rb

@@ -1,4 +1,5 @@
 # frozen_string_literal: true
+
 if ENV['COVERAGE']
   require 'simplecov'
   SimpleCov.start
@@ -60,9 +61,9 @@ class TestFiniteMDP < MiniTest::Test
     assert_equal @r_wait,   model.reward(:high, :wait, :high)
 
     if sparse
-      assert_equal     nil, model.reward(:low, :wait, :high)
-      assert_equal     nil, model.reward(:low, :recharge, :low)
-      assert_equal     nil, model.reward(:high, :wait, :low)
+      assert_nil model.reward(:low, :wait, :high)
+      assert_nil model.reward(:low, :recharge, :low)
+      assert_nil model.reward(:high, :wait, :low)
     else
       assert_equal @r_wait, model.reward(:low, :wait, :high)
       assert_equal 0,       model.reward(:low, :recharge, :low)
@@ -168,11 +169,11 @@ class TestFiniteMDP < MiniTest::Test
     q_low_recharge = 0 + gamma * v[:high]
 
     q = solver.state_action_value
-    assert_close q[[:high, :search]],  q_high_search
-    assert_close q[[:high, :wait]],    q_high_wait
-    assert_close q[[:low, :search]],   q_low_search
-    assert_close q[[:low, :wait]],     q_low_wait
-    assert_close q[[:low, :recharge]], q_low_recharge
+    assert_close q[%i[high search]],  q_high_search
+    assert_close q[%i[high wait]],    q_high_wait
+    assert_close q[%i[low search]],   q_low_search
+    assert_close q[%i[low wait]],     q_low_wait
+    assert_close q[%i[low recharge]], q_low_recharge
   end
 
   #
@@ -314,7 +315,7 @@ class TestFiniteMDP < MiniTest::Test
       [1, 0],         [1, 2], [1, 3],
       [2, 0], [2, 1], [2, 2], [2, 3], :stop], Set[*model.states]
 
-    assert_equal Set[%w(^ > v <)], Set[model.actions([0, 0])]
+    assert_equal Set[%w[^ > v <]], Set[model.actions([0, 0])]
     assert_equal [:stop], model.actions([1, 3])
     assert_equal [:stop], model.actions(:stop)
 
@@ -332,11 +333,11 @@ class TestFiniteMDP < MiniTest::Test
     ], model.hash_to_grid(solver.policy)
 
     # check values against Figure 17.3
-    assert [[0.812, 0.868, 0.918, 1],
+    assert([[0.812, 0.868, 0.918, 1],
             [0.762, nil,   0.660, -1],
             [0.705, 0.655, 0.611, 0.388]].flatten
       .zip(model.hash_to_grid(solver.value).flatten)
-      .all? { |x, y| (x.nil? && y.nil?) || (x - y).abs < 5e-4 }
+      .all? { |x, y| (x.nil? && y.nil?) || (x - y).abs < 5e-4 })
   end
 
   def test_aima_grid_2

metadata

@@ -1,14 +1,14 @@
 --- !ruby/object:Gem::Specification
 name: finite_mdp
 version: !ruby/object:Gem::Version
-  version: 0.3.0
+  version: 0.4.0
 platform: ruby
 authors:
 - John Lees-Miller
 autorequire: 
 bindir: bin
 cert_chain: []
-date: 2016-10-09 00:00:00.000000000 Z
+date: 2017-12-18 00:00:00.000000000 Z
 dependencies:
 - !ruby/object:Gem::Dependency
   name: narray
@@ -44,14 +44,14 @@ dependencies:
     requirements:
     - - "~>"
       - !ruby/object:Gem::Version
-        version: 0.43.0
+        version: 0.52.0
   type: :development
   prerelease: false
   version_requirements: !ruby/object:Gem::Requirement
     requirements:
     - - "~>"
       - !ruby/object:Gem::Version
-        version: 0.43.0
+        version: 0.52.0
 - !ruby/object:Gem::Dependency
   name: simplecov
   requirement: !ruby/object:Gem::Requirement
@@ -96,7 +96,7 @@ rdoc_options:
 - "--main"
 - README.rdoc
 - "--title"
-- finite_mdp-0.3.0 Documentation
+- finite_mdp-0.4.0 Documentation
 require_paths:
 - lib
 required_ruby_version: !ruby/object:Gem::Requirement
@@ -111,7 +111,7 @@ required_rubygems_version: !ruby/object:Gem::Requirement
       version: '0'
 requirements: []
 rubyforge_project: finite_mdp
-rubygems_version: 2.5.1
+rubygems_version: 2.6.13
 signing_key: 
 specification_version: 4
 summary: Solve small, finite Markov Decision Process models.