Changeset 258

Timestamp:

10/15/11 01:41:55 (13 years ago)

Author:

sherbold

Message:

• added methods getChildren() and setCount() to de.ugoe.cs.eventbench.models.TrieNode?
• added method updateKnownSymbols() to de.ugoe.cs.eventbench.models.Trie
• changed the visibility of some members of de.ugoe.cs.eventbench.models.PredictionByPartionMatch?
• added class de.ugoe.cs.eventbench.models.ModelFlattener? to generate FirstOrderMarkovModel? from higher-order models through state splitting
• tweaked entropy calculation of de.ugoe.cs.eventbench.models.FirstOrderMarkovModel? to work with flattened models (locating of START and END was a problem)

Location:

trunk/EventBenchCore/src/de/ugoe/cs/eventbench/models

Files:

• FirstOrderMarkovModel.java (modified) (4 diffs)
• ModelFlattener.java (added)
• PredictionByPartialMatch.java (modified) (2 diffs)
• Trie.java (modified) (2 diffs)
• TrieNode.java (modified) (5 diffs)

trunk/EventBenchCore/src/de/ugoe/cs/eventbench/models/FirstOrderMarkovModel.java

@@ -3 +3 @@
 import java.util.ArrayList;
 import java.util.Collection;
+import java.util.LinkedList;
 import java.util.List;
 import java.util.Random;
@@ -99 +100 @@
                 int numStates = knownSymbols.size();
 
-                int startStateIndex = knownSymbols.indexOf(Event.STARTEVENT);
-                int endStateIndex = knownSymbols.indexOf(Event.ENDEVENT);
-                if (startStateIndex == -1) {
+                List<Integer> startIndexList = new LinkedList<Integer>();
+                List<Integer> endIndexList = new LinkedList<Integer>();
+                for( int i=0 ; i<knownSymbols.size() ; i++ ) {
+                        String id = knownSymbols.get(i).getStandardId();
+                        if( id.equals(Event.STARTEVENT.getStandardId()) || id.contains(Event.STARTEVENT.getStandardId()+"-=-") ) {
+                                startIndexList.add(i);
+                        }
+                        if( id.equals(Event.ENDEVENT.getStandardId()) || id.contains("-=-"+Event.ENDEVENT.getStandardId()) ) {
+                                endIndexList.add(i);
+                        }
+                }
+                
+                if (startIndexList.isEmpty()) { 
                         Console.printerrln("Error calculating entropy. Initial state of markov chain not found.");
                         return Double.NaN;
                 }
-                if (endStateIndex == -1) {
+                if (endIndexList.isEmpty()) {
                         Console.printerrln("Error calculating entropy. End state of markov chain not found.");
                         return Double.NaN;
                 }
-                transmissionMatrix.set(endStateIndex, startStateIndex, 1);
+                for( Integer i : endIndexList ) {
+                        for(Integer j : startIndexList ) {
+                                transmissionMatrix.set(i, j, 1);
+                        }
+                }
 
                 // Calculate stationary distribution by raising the power of the
@@ -160 +175 @@
                 List<Event<?>> knownSymbols = new ArrayList<Event<?>>(
                                 trie.getKnownSymbols());
-
                 for (Event<?> symbol : knownSymbols) {
                         final String thisSaneId = symbol.getShortId().replace("\"", "\\\"")
@@ -183 +197 @@
         /**
          * <p>
-         * Returns a {@link Graph} represenation of the model with the states as
+         * Returns a {@link Graph} representation of the model with the states as
          * nodes and directed edges weighted with transition probabilities.
          * </p>

trunk/EventBenchCore/src/de/ugoe/cs/eventbench/models/PredictionByPartialMatch.java

@@ -34 +34 @@
          * </p>
          */
-        private int minOrder;
+        protected int minOrder;
 
         /**
@@ -41 +41 @@
          * </p>
          */
-        double probEscape;
+        protected double probEscape;
 
         /**

trunk/EventBenchCore/src/de/ugoe/cs/eventbench/models/Trie.java

@@ -85 +85 @@
          */
         public void train(List<T> sequence, int maxOrder) {
-                if( maxOrder<1 ) {
+                if (maxOrder < 1) {
                         return;
                 }
@@ -400 +400 @@
                 return rootNode.getNumLeafs();
         }
+
+        /**
+         * <p>
+         * Updates the list of known symbols by replacing it with all symbols that
+         * are found in the child nodes of the root node. This should be the same as
+         * all symbols that are contained in the trie.
+         * </p>
+         */
+        public void updateKnownSymbols() {
+                knownSymbols = new HashSet<T>();
+                for (TrieNode<T> node : rootNode.getChildren()) {
+                        knownSymbols.add(node.getSymbol());
+                }
+        }
 }

trunk/EventBenchCore/src/de/ugoe/cs/eventbench/models/TrieNode.java

@@ -178 +178 @@
         /**
          * <p>
+         * Returns all children of this node.
+         * </p>
+         * 
+         * @return children of this node
+         */
+        protected Collection<TrieNode<T>> getChildren() {
+                return children;
+        }
+
+        /**
+         * <p>
          * Searches the sub-trie of this trie node for a given sequence and returns
          * the node associated with the sequence or null if no such node is found.
@@ -298 +309 @@
                 return children.isEmpty();
         }
-        
+
         /**
          * <p>
@@ -307 +318 @@
          */
         protected boolean isRoot() {
-                return symbol==null;
+                return symbol == null;
         }
 
@@ -329 +340 @@
                 }
         }
-        
-        /**
-         * <p>Returns the number of descendants of this node that are leafs. This does not only include direct children of this node, but all leafs in the sub-trie with this node as root.
-         * </p>
-         * @return
+
+        /**
+         * <p>
+         * Returns the number of descendants of this node that are leafs. This does
+         * not only include direct children of this node, but all leafs in the
+         * sub-trie with this node as root.
+         * </p>
+         * 
+         * @return number of leafs in this sub-trie
          */
         protected int getNumLeafs() {
                 int numLeafs = 0;
-                for( TrieNode<T> child : children) {
-                        if( child.isLeaf() ) {
+                for (TrieNode<T> child : children) {
+                        if (child.isLeaf()) {
                                 numLeafs++;
                         } else {
@@ -346 +361 @@
                 return numLeafs;
         }
+
+        /**
+         * <p>
+         * Sets the {@link #count} of this node.
+         * </p>
+         * <p>
+         * This function should only be used sparingly and very carefully! The count
+         * is usually maintained automatically by the training procedures.
+         * </p>
+         * 
+         * @param count
+         *            new count
+         */
+        protected void setCount(int count) {
+                this.count = count;
+        }
 }