Thread: [DL-Learner SVN] SF.net SVN: dl-learner:[3529] trunk/components-core/src/main/java/org/ dllearner/

SourceForge Headquarters 1320 Columbia Street Suite 310 San Diego, CA 92101 +1 (858) 422-6466

Revision: 3529
          http://dl-learner.svn.sourceforge.net/dl-learner/?rev=3529&view=rev
Author:   lorenz_b
Date:     2012-01-05 15:15:28 +0000 (Thu, 05 Jan 2012)
Log Message:
-----------
Added some tests for CELOE.

Added Paths:
-----------
    trunk/components-core/src/main/java/org/dllearner/algorithms/celoe/PCELOE.java

Added: trunk/components-core/src/main/java/org/dllearner/algorithms/celoe/PCELOE.java
===================================================================

--- trunk/components-core/src/main/java/org/dllearner/algorithms/celoe/PCELOE.java	                        (rev 0)
+++ trunk/components-core/src/main/java/org/dllearner/algorithms/celoe/PCELOE.java	2012-01-05 15:15:28 UTC (rev 3529)
@@ -0,0 +1,1048 @@
+/**
+ * Copyright (C) 2007-2011, Jens Lehmann
+ *
+ * This file is part of DL-Learner.
+ *
+ * DL-Learner is free software; you can redistribute it and/or modify
+ * it under the terms of the GNU General Public License as published by
+ * the Free Software Foundation; either version 3 of the License, or
+ * (at your option) any later version.
+ *
+ * DL-Learner is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+ * GNU General Public License for more details.
+ *
+ * You should have received a copy of the GNU General Public License
+ * along with this program.  If not, see <http://www.gnu.org/licenses/>.
+ */
+
+package org.dllearner.algorithms.celoe;
+
+import java.io.File;
+import java.text.DecimalFormat;
+import java.util.Collection;
+import java.util.Collections;
+import java.util.HashSet;
+import java.util.Iterator;
+import java.util.LinkedList;
+import java.util.List;
+import java.util.Map;
+import java.util.Set;
+import java.util.SortedSet;
+import java.util.TreeSet;
+import java.util.concurrent.ArrayBlockingQueue;
+import java.util.concurrent.ExecutorService;
+import java.util.concurrent.Executors;
+import java.util.concurrent.ThreadPoolExecutor;
+import java.util.concurrent.TimeUnit;
+
+import org.apache.log4j.Logger;
+import org.dllearner.core.AbstractCELA;
+import org.dllearner.core.AbstractKnowledgeSource;
+import org.dllearner.core.AbstractLearningProblem;
+import org.dllearner.core.AbstractReasonerComponent;
+import org.dllearner.core.ComponentAnn;
+import org.dllearner.core.ComponentInitException;
+import org.dllearner.core.EvaluatedDescription;
+import org.dllearner.core.config.ConfigOption;
+import org.dllearner.core.owl.ClassHierarchy;
+import org.dllearner.core.owl.Description;
+import org.dllearner.core.owl.Individual;
+import org.dllearner.core.owl.Intersection;
+import org.dllearner.core.owl.NamedClass;
+import org.dllearner.core.owl.Restriction;
+import org.dllearner.core.owl.Thing;
+import org.dllearner.kb.OWLFile;
+import org.dllearner.learningproblems.ClassLearningProblem;
+import org.dllearner.learningproblems.PosNegLP;
+import org.dllearner.learningproblems.PosNegLPStandard;
+import org.dllearner.learningproblems.PosOnlyLP;
+import org.dllearner.reasoning.FastInstanceChecker;
+import org.dllearner.refinementoperators.OperatorInverter;
+import org.dllearner.refinementoperators.RefinementOperator;
+import org.dllearner.refinementoperators.RhoDRDown;
+import org.dllearner.utilities.Files;
+import org.dllearner.utilities.Helper;
+import org.dllearner.utilities.owl.ConceptComparator;
+import org.dllearner.utilities.owl.ConceptTransformation;
+import org.dllearner.utilities.owl.DescriptionMinimizer;
+import org.dllearner.utilities.owl.EvaluatedDescriptionSet;
+import org.dllearner.utilities.owl.PropertyContext;
+import org.springframework.beans.factory.annotation.Autowired;
+
+import com.jamonapi.Monitor;
+import com.jamonapi.MonitorFactory;
+
+/**
+ * The CELOE (Class Expression Learner for Ontology Engineering) algorithm.
+ * It adapts and extends the standard supervised learning algorithm for the
+ * ontology engineering use case. 
+ * 
+ * @author Jens Lehmann
+ *
+ */
+@ComponentAnn(name="PCELOE", shortName="pceloe", version=1.0, description="CELOE is an adapted and extended version of the OCEL algorithm applied for the ontology engineering use case. See http://jens-lehmann.org/files/2011/celoe.pdf for reference.")
+public class PCELOE extends AbstractCELA {
+	
+	//parameters for thread pool
+		//Parallel running Threads(Executor) on System
+		private static int corePoolSize = 5;
+		//Maximum Threads allowed in Pool
+		private static int maximumPoolSize = 20;
+		//Keep alive time for waiting threads for jobs(Runnable)
+		private static long keepAliveTime = 10;
+	
+
+	private static Logger logger = Logger.getLogger(PCELOE.class);
+//	private CELOEConfigurator configurator;
+	
+	private boolean isRunning = false;
+	private boolean stop = false;	
+	
+//	private OEHeuristicStable heuristicStable = new OEHeuristicStable();
+//	private OEHeuristicRuntime heuristicRuntime = new OEHeuristicRuntime();
+	
+	private RefinementOperator operator;
+	private DescriptionMinimizer minimizer;
+	@ConfigOption(name="useMinimizer", defaultValue="true", description="Specifies whether returned expressions should be minimised by removing those parts, which are not needed. (Basically the minimiser tries to find the shortest expression which is equivalent to the learned expression). Turning this feature off may improve performance.")
+	private boolean useMinimizer = true;
+	
+	// all nodes in the search tree (used for selecting most promising node)
+	private TreeSet<OENode> nodes;
+	private OEHeuristicRuntime heuristic; // = new OEHeuristicRuntime();
+	// root of search tree
+	private OENode startNode;
+	// the class with which we start the refinement process
+	@ConfigOption(name = "startClass", defaultValue="owl:Thing", description="You can specify a start class for the algorithm. To do this, you have to use Manchester OWL syntax without using prefixes.")
+	private Description startClass;
+	
+	// all descriptions in the search tree plus those which were too weak (for fast redundancy check)
+	private TreeSet<Description> descriptions;
+	
+	private EvaluatedDescriptionSet bestEvaluatedDescriptions;
+	
+	// if true, then each solution is evaluated exactly instead of approximately
+	// private boolean exactBestDescriptionEvaluation = false;
+	@ConfigOption(name = "singleSuggestionMode", defaultValue="false", description="Use this if you are interested in only one suggestion and your learning problem has many (more than 1000) examples.")
+	private boolean singleSuggestionMode;
+	private Description bestDescription;
+	private double bestAccuracy = Double.MIN_VALUE;
+	
+	private NamedClass classToDescribe;
+	// examples are either 1.) instances of the class to describe 2.) positive examples
+	// 3.) union of pos.+neg. examples depending on the learning problem at hand
+	private Set<Individual> examples;
+	
+	// CELOE was originally created for learning classes in ontologies, but also
+	// works for other learning problem types
+	private boolean isClassLearningProblem;
+	private boolean isEquivalenceProblem;
+	
+	private long nanoStartTime;
+	
+	// important parameters (non-config options but internal)
+	private double noise;
+
+	private boolean filterFollowsFromKB;	
+	
+	// less important parameters
+	// forces that one solution cannot be subexpression of another expression; this option is useful to get diversity
+	// but it can also suppress quite useful expressions
+	private boolean forceMutualDifference = false;
+	
+	// utility variables
+	private String baseURI;
+	private Map<String, String> prefixes;
+	private DecimalFormat dfPercent = new DecimalFormat("0.00%");
+	private ConceptComparator descriptionComparator = new ConceptComparator();
+	
+	// statistical variables
+	private int expressionTests = 0;
+	private int minHorizExp = 0;
+	private int maxHorizExp = 0;
+	
+	// TODO: turn those into config options
+	
+	// important: do not initialise those with empty sets
+	// null = no settings for allowance / ignorance
+	// empty set = allow / ignore nothing (it is often not desired to allow no class!)
+	Set<NamedClass> allowedConcepts = null;
+	Set<NamedClass> ignoredConcepts = null;
+
+	@ConfigOption(name = "writeSearchTree", defaultValue="false", description="specifies whether to write a search tree")
+	private boolean writeSearchTree = false;
+
+	@ConfigOption(name = "searchTreeFile", defaultValue="log/searchTree.txt", description="file to use for the search tree")
+	private String searchTreeFile = "log/searchTree.txt";
+
+	@ConfigOption(name = "replaceSearchTree", defaultValue="false", description="specifies whether to replace the search tree in the log file after each run or append the new search tree")
+	private boolean replaceSearchTree = false;
+	
+	@ConfigOption(name = "maxNrOfResults", defaultValue="10", description="Sets the maximum number of results one is interested in. (Setting this to a lower value may increase performance as the learning algorithm has to store/evaluate/beautify less descriptions).")	
+	private int maxNrOfResults = 10;
+
+	@ConfigOption(name = "noisePercentage", defaultValue="0.0", description="the (approximated) percentage of noise within the examples")
+	private double noisePercentage = 0.0;
+
+	@ConfigOption(name = "filterDescriptionsFollowingFromKB", defaultValue="false", description="If true, then the results will not contain suggestions, which already follow logically from the knowledge base. Be careful, since this requires a potentially expensive consistency check for candidate solutions.")
+	private boolean filterDescriptionsFollowingFromKB = false;
+
+	@ConfigOption(name = "reuseExistingDescription", defaultValue="false", description="If true, the algorithm tries to find a good starting point close to an existing definition/super class of the given class in the knowledge base.")
+	private boolean reuseExistingDescription = false;
+
+	@ConfigOption(name = "maxClassDescriptionTests", defaultValue="0", description="The maximum number of candidate hypothesis the algorithm is allowed to test (0 = no limit). The algorithm will stop afterwards. (The real number of tests can be slightly higher, because this criterion usually won't be checked after each single test.)")
+	private int maxClassDescriptionTests = 0;
+
+	@ConfigOption(defaultValue = "10", name = "maxExecutionTimeInSeconds", description = "maximum execution of the algorithm in seconds")
+	private int maxExecutionTimeInSeconds = 10;
+
+	@ConfigOption(name = "terminateOnNoiseReached", defaultValue="false", description="specifies whether to terminate when noise criterion is met")
+	private boolean terminateOnNoiseReached = false;
+	
+	@ConfigOption(name = "maxDepth", defaultValue="7", description="maximum depth of description")
+	private double maxDepth = 7;
+	
+	private Set<OENode> currentlyProcessedNodes = Collections.synchronizedSet(new HashSet<OENode>());
+	
+//	public CELOEConfigurator getConfigurator() {
+//		return configurator;
+//	}
+	
+	public PCELOE() {
+		
+	}
+	
+	public PCELOE(AbstractLearningProblem problem, AbstractReasonerComponent reasoner) {
+		super(problem, reasoner);
+//		configurator = new CELOEConfigurator(this);
+	}
+
+//	public static Collection<Class<? extends AbstractLearningProblem>> supportedLearningProblems() {
+//		Collection<Class<? extends AbstractLearningProblem>> problems = new LinkedList<Class<? extends AbstractLearningProblem>>();
+//		problems.add(AbstractLearningProblem.class);
+//		return problems;
+//	}
+	
+	public static String getName() {
+		return "CELOE";
+	}
+	
+	@Override
+	public void init() throws ComponentInitException {
+			
+		// compute used concepts/roles from allowed/ignored
+		// concepts/roles
+		Set<NamedClass> usedConcepts;
+//		Set<NamedClass> allowedConcepts = configurator.getAllowedConcepts()==null ? null : CommonConfigMappings.getAtomicConceptSet(configurator.getAllowedConcepts());
+//		Set<NamedClass> ignoredConcepts = configurator.getIgnoredConcepts()==null ? null : CommonConfigMappings.getAtomicConceptSet(configurator.getIgnoredConcepts());
+		if(allowedConcepts != null) {
+			// sanity check to control if no non-existing concepts are in the list
+			Helper.checkConcepts(reasoner, allowedConcepts);
+			usedConcepts = allowedConcepts;
+		} else if(ignoredConcepts != null) {
+			usedConcepts = Helper.computeConceptsUsingIgnoreList(reasoner, ignoredConcepts);
+		} else {
+			usedConcepts = Helper.computeConcepts(reasoner);
+		}
+		
+		// copy class hierarchy and modify it such that each class is only
+		// reachable via a single path
+//		ClassHierarchy classHierarchy = reasoner.getClassHierarchy().clone();
+		ClassHierarchy classHierarchy = reasoner.getClassHierarchy().cloneAndRestrict(usedConcepts);
+		classHierarchy.thinOutSubsumptionHierarchy();
+
+		// if no one injected a heuristic, we use a default one
+		if(heuristic == null) {
+			heuristic = new OEHeuristicRuntime();
+		}
+		
+		minimizer = new DescriptionMinimizer(reasoner);
+		
+		// start at owl:Thing by default
+		if(startClass == null) {
+			startClass = Thing.instance;
+		}
+		
+//		singleSuggestionMode = configurator.getSingleSuggestionMode();
+		/*
+		// create refinement operator
+		if(operator == null) {
+			operator = new RhoDRDown();
+			((RhoDRDown)operator).setStartClass(startClass);
+			((RhoDRDown)operator).setReasoner(reasoner);
+		}
+		((RhoDRDown)operator).setSubHierarchy(classHierarchy);
+		((RhoDRDown)operator).setObjectPropertyHierarchy(reasoner.getObjectPropertyHierarchy());
+		((RhoDRDown)operator).setDataPropertyHierarchy(reasoner.getDatatypePropertyHierarchy());
+		((RhoDRDown)operator).init();		
+		*/
+		// create a refinement operator and pass all configuration
+		// variables to it
+		if(operator == null) {
+			// we use a default operator and inject the class hierarchy for now
+			operator = new RhoDRDown();
+			((RhoDRDown)operator).setStartClass(startClass);
+			((RhoDRDown)operator).setReasoner(reasoner);
+			((RhoDRDown)operator).init();
+		}
+		// TODO: find a better solution as this is quite difficult to debug
+		((RhoDRDown)operator).setSubHierarchy(classHierarchy);
+		((RhoDRDown)operator).setObjectPropertyHierarchy(reasoner.getObjectPropertyHierarchy());
+		((RhoDRDown)operator).setDataPropertyHierarchy(reasoner.getDatatypePropertyHierarchy());
+		
+		
+//		operator = new RhoDRDown(reasoner, classHierarchy, startClass, configurator);
+		baseURI = reasoner.getBaseURI();
+		prefixes = reasoner.getPrefixes();		
+		if(writeSearchTree) {
+			File f = new File(searchTreeFile );
+			Files.clearFile(f);
+		}
+		
+		bestEvaluatedDescriptions = new EvaluatedDescriptionSet(maxNrOfResults);
+		
+		isClassLearningProblem = (learningProblem instanceof ClassLearningProblem);
+		
+		// we put important parameters in class variables
+		noise = noisePercentage/100d;
+//		System.out.println("noise " + noise);
+//		maxDepth = configurator.getMaxDepth();
+		// (filterFollowsFromKB is automatically set to false if the problem
+		// is not a class learning problem
+		filterFollowsFromKB = filterDescriptionsFollowingFromKB && isClassLearningProblem;
+		
+//		Set<Description> concepts = operator.refine(Thing.instance, 5);
+//		for(Description concept : concepts) {
+//			System.out.println(concept);
+//		}
+//		System.out.println("refinements of thing: " + concepts.size());
+		
+		// actions specific to ontology engineering
+		if(isClassLearningProblem) {
+			ClassLearningProblem problem = (ClassLearningProblem) learningProblem;
+			classToDescribe = problem.getClassToDescribe();
+			isEquivalenceProblem = problem.isEquivalenceProblem();
+			
+			examples = reasoner.getIndividuals(classToDescribe);
+			
+			// start class: intersection of super classes for definitions (since it needs to
+			// capture all instances), but owl:Thing for learning subclasses (since it is
+			// superfluous to add super classes in this case)
+			if(isEquivalenceProblem) {
+				Set<Description> existingDefinitions = reasoner.getAssertedDefinitions(classToDescribe);
+				if(reuseExistingDescription && (existingDefinitions.size() > 0)) {
+					// the existing definition is reused, which in the simplest case means to
+					// use it as a start class or, if it is already too specific, generalise it
+					
+					// pick the longest existing definition as candidate
+					Description existingDefinition = null;
+					int highestLength = 0;
+					for(Description exDef : existingDefinitions) {
+						if(exDef.getLength() > highestLength) {
+							existingDefinition = exDef;
+							highestLength = exDef.getLength();
+						}
+					}
+					
+					LinkedList<Description> startClassCandidates = new LinkedList<Description>();
+					startClassCandidates.add(existingDefinition);
+					((RhoDRDown)operator).setDropDisjuncts(true);
+					RefinementOperator upwardOperator = new OperatorInverter(operator);
+					
+					// use upward refinement until we find an appropriate start class
+					boolean startClassFound = false;
+					Description candidate;
+					do {
+						candidate = startClassCandidates.pollFirst();
+						if(((ClassLearningProblem)learningProblem).getRecall(candidate)<1.0) {
+							// add upward refinements to list
+							Set<Description> refinements = upwardOperator.refine(candidate, candidate.getLength());
+//							System.out.println("ref: " + refinements);
+							LinkedList<Description> refinementList = new LinkedList<Description>(refinements);
+//							Collections.reverse(refinementList);
+//							System.out.println("list: " + refinementList);
+							startClassCandidates.addAll(refinementList);
+//							System.out.println("candidates: " + startClassCandidates);
+						} else {
+							startClassFound = true;
+						}
+					} while(!startClassFound);
+					startClass = candidate;
+					
+					if(startClass.equals(existingDefinition)) {
+						logger.info("Reusing existing description " + startClass.toManchesterSyntaxString(baseURI, prefixes) + " as start class for learning algorithm.");
+					} else {
+						logger.info("Generalised existing description " + existingDefinition.toManchesterSyntaxString(baseURI, prefixes) + " to " + startClass.toManchesterSyntaxString(baseURI, prefixes) + ", which is used as start class for the learning algorithm.");
+					}
+					
+//					System.out.println("start class: " + startClass);
+//					System.out.println("existing def: " + existingDefinition);
+//					System.out.println(reasoner.getIndividuals(existingDefinition));
+					
+					((RhoDRDown)operator).setDropDisjuncts(false);
+					
+				} else {
+					Set<Description> superClasses = reasoner.getClassHierarchy().getSuperClasses(classToDescribe);
+					if(superClasses.size() > 1) {
+						startClass = new Intersection(new LinkedList<Description>(superClasses));
+					} else if(superClasses.size() == 1){
+						startClass = (Description) superClasses.toArray()[0];
+					} else {
+						startClass = Thing.instance;
+						logger.warn(classToDescribe + " is equivalent to owl:Thing. Usually, it is not " +
+								"sensible to learn a description in this case.");
+					}					
+				}
+			}				
+		} else if(learningProblem instanceof PosOnlyLP) {
+			examples = ((PosOnlyLP)learningProblem).getPositiveExamples();
+		} else if(learningProblem instanceof PosNegLP) {
+			examples = Helper.union(((PosNegLP)learningProblem).getPositiveExamples(),((PosNegLP)learningProblem).getNegativeExamples());
+		}
+	}
+
+	@Override
+	public Description getCurrentlyBestDescription() {
+		EvaluatedDescription ed = getCurrentlyBestEvaluatedDescription();
+		return ed == null ? null : ed.getDescription();
+	}
+
+	@Override
+	public List<Description> getCurrentlyBestDescriptions() {
+		return bestEvaluatedDescriptions.toDescriptionList();
+	}
+	
+	@Override
+	public EvaluatedDescription getCurrentlyBestEvaluatedDescription() {
+		return bestEvaluatedDescriptions.getBest();
+	}	
+	
+	@Override
+	public TreeSet<? extends EvaluatedDescription> getCurrentlyBestEvaluatedDescriptions() {
+		return bestEvaluatedDescriptions.getSet();
+	}	
+	
+	public double getCurrentlyBestAccuracy() {
+		return bestEvaluatedDescriptions.getBest().getAccuracy();
+	}
+	
+	@Override
+	public void start() {
+//		System.out.println(configurator.getMaxExecutionTimeInSeconds());
+		
+		stop = false;
+		isRunning = true;
+		reset();
+		nanoStartTime = System.nanoTime();
+		
+		addNode(startClass, null);
+		
+		int nrOfThreads = Runtime.getRuntime().availableProcessors();
+		ExecutorService service = Executors.newFixedThreadPool(nrOfThreads);
+	
+		for(int i = 0; i < 2; i++){
+			service.submit(new Runnable() {
+				
+				@Override
+				public void run() {
+					OENode nextNode;
+					double highestAccuracy = 0.0;
+					int loop = 0;
+					while (!terminationCriteriaSatisfied()) {
+						
+						
+						if(!singleSuggestionMode && bestEvaluatedDescriptions.getBestAccuracy() > highestAccuracy) {
+							highestAccuracy = bestEvaluatedDescriptions.getBestAccuracy();
+							logger.info("more accurate (" + dfPercent.format(highestAccuracy) + ") class expression found: " + descriptionToString(bestEvaluatedDescriptions.getBest().getDescription()));	
+						}
+
+						// chose best node according to heuristics
+						nextNode = getNextNodeToExpand();
+						try {
+							Thread.sleep(10);
+						} catch (InterruptedException e) {
+							// TODO Auto-generated catch block
+							e.printStackTrace();
+						}
+						logger.info("next Node: " + nextNode);
+						if(nextNode != null){
+							int horizExp = nextNode.getHorizontalExpansion();
+							
+							// apply operator
+							Monitor mon = MonitorFactory.start("refineNode");
+							TreeSet<Description> refinements = refineNode(nextNode);
+							mon.stop();
+							
+							while(refinements.size() != 0) {
+								// pick element from set
+								Description refinement = refinements.pollFirst();
+								int length = refinement.getLength();
+												
+								// we ignore all refinements with lower length and too high depth
+								// (this also avoids duplicate node children)
+								if(length > horizExp && refinement.getDepth() <= maxDepth) {
+									
+//									System.out.println("potentially adding " + refinement + " to search tree as child of " + nextNode + " " + new Date());
+									Monitor mon2 = MonitorFactory.start("addNode");
+									addNode(refinement, nextNode);
+									mon2.stop();
+									// adding nodes is potentially computationally expensive, so we have
+									// to check whether max time is exceeded	
+									if(terminationCriteriaSatisfied()) {
+										break;
+									}
+//									System.out.println("addNode finished" + " " + new Date());
+								}
+						
+//								System.out.println("  refinement queue length: " + refinements.size());
+							}
+							
+//							updateMinMaxHorizExp(nextNode);
+								
+							loop++;
+							currentlyProcessedNodes.remove(nextNode);
+						}
+						
+					}
+					
+				}
+			});
+		}
+		
+		try {
+			service.awaitTermination(maxExecutionTimeInSeconds, TimeUnit.SECONDS);
+		} catch (InterruptedException e) {
+			e.printStackTrace();
+		}
+		
+		if(singleSuggestionMode) {
+			bestEvaluatedDescriptions.add(bestDescription, bestAccuracy, learningProblem);
+		}	
+		
+		if (stop) {
+			logger.info("Algorithm stopped ("+expressionTests+" descriptions tested). " + nodes.size() + " nodes in the search tree.\n");
+		} else {
+			logger.info("Algorithm terminated successfully ("+expressionTests+" descriptions tested). "  + nodes.size() + " nodes in the search tree.\n");
+            logger.info(reasoner.toString());
+		}
+		
+		// print solution(s)
+		logger.info("solutions:\n" + getSolutionString());
+		
+//		System.out.println(startNode.toTreeString(baseURI));
+		
+		isRunning = false;
+//		System.out.println("isRunning: " + isRunning);
+		service.shutdown();
+		
+	}
+
+	private OENode getNextNodeToExpand() {
+		// we expand the best node of those, which have not achieved 100% accuracy
+		// already and have a horizontal expansion equal to their length
+		// (rationale: further extension is likely to add irrelevant syntactical constructs)
+		Iterator<OENode> it = nodes.descendingIterator();//logger.info(nodes.size());
+		while(it.hasNext()) {
+			OENode node = it.next();
+			if(!currentlyProcessedNodes.contains(node) && (node.getAccuracy() < 1.0 || node.getHorizontalExpansion() < node.getDescription().getLength())) {
+				currentlyProcessedNodes.add(node);
+				return node;
+			}
+		}
+		
+		// this should practically never be called, since for any reasonable learning
+		// task, we will always have at least one node with less than 100% accuracy
+		return null;
+	}
+	
+	// expand node horizontically
+	private TreeSet<Description> refineNode(OENode node) {
+		// we have to remove and add the node since its heuristic evaluation changes through the expansion
+		// (you *must not* include any criteria in the heuristic which are modified outside of this method,
+		// otherwise you may see rarely occurring but critical false ordering in the nodes set)
+		nodes.remove(node);
+//		System.out.println("refining: " + node);
+		int horizExp = node.getHorizontalExpansion();
+		TreeSet<Description> refinements = (TreeSet<Description>) operator.refine(node.getDescription(), horizExp+1);
+		node.incHorizontalExpansion();
+		node.setRefinementCount(refinements.size());
+		nodes.add(node);
+		return refinements;
+	}
+	
+	// add node to search tree if it is not too weak
+	// returns true if node was added and false otherwise
+	private boolean addNode(Description description, OENode parentNode) {
+		
+//		System.out.println("d: " + description);
+		
+		// redundancy check (return if redundant)
+		boolean nonRedundant = descriptions.add(description);
+		if(!nonRedundant) {
+			return false;
+		}
+		
+		// check whether the description is allowed
+		if(!isDescriptionAllowed(description, parentNode)) {
+			return false;
+		}
+		
+//		System.out.println("Test " + new Date());
+		// quality of description (return if too weak)
+		double accuracy = learningProblem.getAccuracyOrTooWeak(description, noise);
+		// issue a warning if accuracy is not between 0 and 1 or -1 (too weak)
+		if(accuracy > 1.0 || (accuracy < 0.0 && accuracy != -1)) {
+			logger.warn("Invalid accuracy value " + accuracy + " for description " + description + ". This could be caused by a bug in the heuristic measure and should be reported to the DL-Learner bug tracker.");
+			System.exit(0);
+		}
+		
+//		System.out.println("Test2 " + new Date());
+		expressionTests++;
+//		System.out.println("acc: " + accuracy);
+//		System.out.println(description + " " + accuracy);
+		if(accuracy == -1) {
+			return false;
+		}
+		
+		OENode node = new OENode(parentNode, description, accuracy);
+			
+		// link to parent (unless start node)
+		if(parentNode == null) {
+			startNode = node;
+		} else {
+			parentNode.addChild(node);
+		}
+	
+		nodes.add(node);
+//		System.out.println("Test3 " + new Date());
+		
+		// in some cases (e.g. mutation) fully evaluating even a single description is too expensive
+		// due to the high number of examples -- so we just stick to the approximate accuracy
+		if(singleSuggestionMode) {
+			if(accuracy > bestAccuracy) {
+				bestAccuracy = accuracy;
+				bestDescription = description;
+				logger.info("more accurate (" + dfPercent.format(bestAccuracy) + ") class expression found: " + descriptionToString(bestDescription)); // + getTemporaryString(bestDescription)); 
+			}
+			return true;
+		} 
+		
+//		System.out.println("description " + description + " accuracy " + accuracy);
+		
+		// maybe add to best descriptions (method keeps set size fixed);
+		// we need to make sure that this does not get called more often than
+		// necessary since rewriting is expensive
+		boolean isCandidate = !bestEvaluatedDescriptions.isFull();
+		if(!isCandidate) {
+			EvaluatedDescription worst = bestEvaluatedDescriptions.getWorst();
+			double accThreshold = worst.getAccuracy();
+			isCandidate = 
+				(accuracy > accThreshold ||
+				(accuracy >= accThreshold && description.getLength() < worst.getDescriptionLength()));
+		}
+		
+//		System.out.println(isCandidate);
+		
+//		System.out.println("Test4 " + new Date());
+		if(isCandidate) {
+			
+			Description niceDescription = rewriteNode(node);
+			ConceptTransformation.transformToOrderedForm(niceDescription, descriptionComparator);
+//			Description niceDescription = node.getDescription();
+			
+			// another test: none of the other suggested descriptions should be 
+			// a subdescription of this one unless accuracy is different
+			// => comment: on the one hand, this appears to be too strict, because once A is a solution then everything containing
+			// A is not a candidate; on the other hand this suppresses many meaningless extensions of A
+			boolean shorterDescriptionExists = false;
+			if(forceMutualDifference) {
+				for(EvaluatedDescription ed : bestEvaluatedDescriptions.getSet()) {
+					if(Math.abs(ed.getAccuracy()-accuracy) <= 0.00001 && ConceptTransformation.isSubdescription(niceDescription, ed.getDescription())) {
+//						System.out.println("shorter: " + ed.getDescription());
+						shorterDescriptionExists = true;
+						break;
+					}
+				}				
+			}
+			
+//			System.out.println("shorter description? " + shorterDescriptionExists + " nice: " + niceDescription);
+			
+			if(!shorterDescriptionExists) {
+				if(!filterFollowsFromKB || !((ClassLearningProblem)learningProblem).followsFromKB(niceDescription)) {
+//					System.out.println("Test2");
+					bestEvaluatedDescriptions.add(niceDescription, accuracy, learningProblem);
+//					System.out.println("acc: " + accuracy);
+//					System.out.println(bestEvaluatedDescriptions);
+				}
+			}
+						
+//			System.out.println(bestEvaluatedDescriptions.getSet().size());
+		}
+		
+//		System.out.println("Test5 " + new Date());
+//		System.out.println("best evaluated descriptions size: " + bestEvaluatedDescriptions.size() + " worst: " + bestEvaluatedDescriptions.getWorst());
+		return true;
+	}	
+	
+	// checks whether the description is allowed
+	private boolean isDescriptionAllowed(Description description, OENode parentNode) {
+		if(isClassLearningProblem) {
+			if(isEquivalenceProblem) {
+				// the class to learn must not appear on the outermost property level
+				if(occursOnFirstLevel(description, classToDescribe)) {
+					return false;
+				}
+			} else {
+				// none of the superclasses of the class to learn must appear on the
+				// outermost property level
+				TreeSet<Description> toTest = new TreeSet<Description>(descriptionComparator);
+				toTest.add(classToDescribe);
+				while(!toTest.isEmpty()) {
+					Description d = toTest.pollFirst();
+					if(occursOnFirstLevel(description, d)) {
+						return false;
+					}
+					toTest.addAll(reasoner.getClassHierarchy().getSuperClasses(d));
+				}
+			}			
+		}
+		
+		// perform forall sanity tests
+		if(parentNode != null && ConceptTransformation.getForallOccurences(description) > ConceptTransformation.getForallOccurences(parentNode.getDescription())) {
+			// we have an additional \forall construct, so we now fetch the contexts
+			// in which it occurs
+			SortedSet<PropertyContext> contexts = ConceptTransformation.getForallContexts(description);
+			SortedSet<PropertyContext> parentContexts = ConceptTransformation.getForallContexts(parentNode.getDescription());
+			contexts.removeAll(parentContexts);
+//			System.out.println("parent description: " + parentNode.getDescription());
+//			System.out.println("description: " + description);
+//			System.out.println("contexts: " + contexts);
+			// we now have to perform sanity checks: if \forall is used, then there
+			// should be at least on class instance which has a filler at the given context
+			for(PropertyContext context : contexts) {
+				// transform [r,s] to \exists r.\exists s.\top
+				Description existentialContext = context.toExistentialContext();
+				boolean fillerFound = false;
+				for(Individual instance : examples) {
+					if(reasoner.hasType(existentialContext, instance)) {
+//						System.out.println(instance + "  " + existentialContext);
+						fillerFound = true;
+						break;
+					}
+				}
+				// if we do not find a filler, this means that putting \forall at
+				// that position is not meaningful
+				if(!fillerFound) {
+					return false;
+				}
+			}
+		}		
+		
+		// we do not want to have negations of sibling classes on the outermost level
+		// (they are expressed more naturally by saying that the siblings are disjoint,
+		// so it is reasonable not to include them in solutions)
+//		Set<Description> siblingClasses = reasoner.getClassHierarchy().getSiblingClasses(classToDescribe);
+//		for now, we just disable negation
+		
+		return true;
+	}
+	
+	// determine whether a named class occurs on the outermost level, i.e. property depth 0
+	// (it can still be at higher depth, e.g. if intersections are nested in unions)
+	private boolean occursOnFirstLevel(Description description, Description clazz) {
+		if(description instanceof NamedClass) {
+			if(description.equals(clazz)) {
+				return true;
+			}
+		} 
+		
+		if(description instanceof Restriction) {
+			return false;
+		}
+		
+		for(Description child : description.getChildren()) {
+			if(occursOnFirstLevel(child, clazz)) {
+				return true;
+			}
+		}
+		
+		return false;
+	}
+	
+	// check whether the node is a potential solution candidate
+	private Description rewriteNode(OENode node) {
+		Description description = node.getDescription();
+		// minimize description (expensive!) - also performes some human friendly rewrites
+		Description niceDescription;
+		if(useMinimizer) {
+			niceDescription = minimizer.minimizeClone(description);
+		} else {
+			niceDescription = description;
+		}
+		// replace \exists r.\top with \exists r.range(r) which is easier to read for humans
+		ConceptTransformation.replaceRange(niceDescription, reasoner);
+		return niceDescription;
+	}
+	
+	private boolean terminationCriteriaSatisfied() {
+		return 
+		stop || 
+		(maxClassDescriptionTests != 0 && (expressionTests >= maxClassDescriptionTests)) ||
+		(maxExecutionTimeInSeconds != 0 && ((System.nanoTime() - nanoStartTime) >= (maxExecutionTimeInSeconds*1000000000l))) ||
+		(terminateOnNoiseReached && (100*getCurrentlyBestAccuracy()>=100-noisePercentage));
+	}
+	
+	private void reset() {
+		// set all values back to their default values (used for running
+		// the algorithm more than once)
+		nodes = new TreeSet<OENode>(heuristic);
+		descriptions = new TreeSet<Description>(new ConceptComparator());
+		bestEvaluatedDescriptions.getSet().clear();
+		expressionTests = 0;
+	}
+	
+	@Override
+	public boolean isRunning() {
+		return isRunning;
+	}	
+	
+	@Override
+	public void stop() {
+		stop = true;
+	}
+
+	public OENode getSearchTreeRoot() {
+		return startNode;
+	}
+	
+	// central function for printing description
+	private String descriptionToString(Description description) {
+		return description.toManchesterSyntaxString(baseURI, prefixes);
+	}
+	
+	@SuppressWarnings("unused")
+	private String bestDescriptionToString() {
+		EvaluatedDescription best = bestEvaluatedDescriptions.getBest();
+		return best.getDescription().toManchesterSyntaxString(baseURI, prefixes) + " (accuracy: " + dfPercent.format(best.getAccuracy()) + ")";
+	}	
+	
+	private String getSolutionString() {
+		int current = 1;
+		String str = "";
+		for(EvaluatedDescription ed : bestEvaluatedDescriptions.getSet().descendingSet()) {
+			// temporary code
+			if(learningProblem instanceof PosNegLPStandard) {
+				str += current + ": " + descriptionToString(ed.getDescription()) + " (pred. acc.: " + dfPercent.format(((PosNegLPStandard)learningProblem).getPredAccuracyOrTooWeakExact(ed.getDescription(),1)) + ", F-measure: "+ dfPercent.format(((PosNegLPStandard)learningProblem).getFMeasureOrTooWeakExact(ed.getDescription(),1)) + ")\n";
+			} else {
+				str += current + ": " + descriptionToString(ed.getDescription()) + " " + dfPercent.format(ed.getAccuracy()) + "\n";
+//				System.out.println(ed);
+			}
+			current++;
+		}
+		return str;
+	}
+
+//	private String getTemporaryString(Description description) {
+//		return descriptionToString(description) + " (pred. acc.: " + dfPercent.format(((PosNegLPStandard)learningProblem).getPredAccuracyOrTooWeakExact(description,1)) + ", F-measure: "+ dfPercent.format(((PosNegLPStandard)learningProblem).getFMeasureOrTooWeakExact(description,1)) + ")";
+//	}
+	
+	private void updateMinMaxHorizExp(OENode node) {
+		int newHorizExp = node.getHorizontalExpansion();
+		
+		// update maximum value
+		maxHorizExp = Math.max(maxHorizExp, newHorizExp);
+		
+		// we just expanded a node with minimum horizontal expansion;
+		// we need to check whether it was the last one
+		if(minHorizExp == newHorizExp - 1) {
+			
+			// the best accuracy that a node can achieve 
+			double scoreThreshold = heuristic.getNodeScore(node) + 1 - node.getAccuracy();
+			
+			for(OENode n : nodes.descendingSet()) {
+				if(n != node) {
+					if(n.getHorizontalExpansion() == minHorizExp) {
+						// we can stop instantly when another node with min. 
+						return;
+					}
+					if(heuristic.getNodeScore(n) < scoreThreshold) {
+						// we can stop traversing nodes when their score is too low
+						break;
+					}
+				}
+			}
+			
+			// inc. minimum since we found no other node which also has min. horiz. exp.
+			minHorizExp++;
+			
+//			System.out.println("minimum horizontal expansion is now " + minHorizExp);
+		}
+	}
+	
+	public int getMaximumHorizontalExpansion() {
+		return maxHorizExp;
+	}
+
+	public int getMinimumHorizontalExpansion() {
+		return minHorizExp;
+	}
+	
+	/**
+	 * @return the expressionTests
+	 */
+	public int getClassExpressionTests() {
+		return expressionTests;
+	}
+
+	public RefinementOperator getOperator() {
+		return operator;
+	}
+
+	@Autowired(required=false)
+	public void setOperator(RefinementOperator operator) {
+		this.operator = operator;
+	}
+
+	public Description getStartClass() {
+		return startClass;
+	}
+
+	public void setStartClass(Description startClass) {
+		this.startClass = startClass;
+	}
+
+	public Set<NamedClass> getAllowedConcepts() {
+		return allowedConcepts;
+	}
+
+	public void setAllowedConcepts(Set<NamedClass> allowedConcepts) {
+		this.allowedConcepts = allowedConcepts;
+	}
+
+	public Set<NamedClass> getIgnoredConcepts() {
+		return ignoredConcepts;
+	}
+
+	public void setIgnoredConcepts(Set<NamedClass> ignoredConcepts) {
+		this.ignoredConcepts = ignoredConcepts;
+	}
+
+	public boolean isWriteSearchTree() {
+		return writeSearchTree;
+	}
+
+	public void setWriteSearchTree(boolean writeSearchTree) {
+		this.writeSearchTree = writeSearchTree;
+	}
+
+	public String getSearchTreeFile() {
+		return searchTreeFile;
+	}
+
+	public void setSearchTreeFile(String searchTreeFile) {
+		this.searchTreeFile = searchTreeFile;
+	}
+
+	public int getMaxNrOfResults() {
+		return maxNrOfResults;
+	}
+
+	public void setMaxNrOfResults(int maxNrOfResults) {
+		this.maxNrOfResults = maxNrOfResults;
+	}
+
+	public double getNoisePercentage() {
+		return noisePercentage;
+	}
+
+	public void setNoisePercentage(double noisePercentage) {
+		this.noisePercentage = noisePercentage;
+	}
+
+	public boolean isFilterDescriptionsFollowingFromKB() {
+		return filterDescriptionsFollowingFromKB;
+	}
+
+	public void setFilterDescriptionsFollowingFromKB(boolean filterDescriptionsFollowingFromKB) {
+		this.filterDescriptionsFollowingFromKB = filterDescriptionsFollowingFromKB;
+	}
+
+	public boolean isReplaceSearchTree() {
+		return replaceSearchTree;
+	}
+
+	public void setReplaceSearchTree(boolean replaceSearchTree) {
+		this.replaceSearchTree = replaceSearchTree;
+	}
+
+	public int getMaxClassDescriptionTests() {
+		return maxClassDescriptionTests;
+	}
+
+	public void setMaxClassDescriptionTests(int maxClassDescriptionTests) {
+		this.maxClassDescriptionTests = maxClassDescriptionTests;
+	}
+
+	public int getMaxExecutionTimeInSeconds() {
+		return maxExecutionTimeInSeconds;
+	}
+
+	public void setMaxExecutionTimeInSeconds(int maxExecutionTimeInSeconds) {
+		this.maxExecutionTimeInSeconds = maxExecutionTimeInSeconds;
+	}
+
+	public boolean isTerminateOnNoiseReached() {
+		return terminateOnNoiseReached;
+	}
+
+	public void setTerminateOnNoiseReached(boolean terminateOnNoiseReached) {
+		this.terminateOnNoiseReached = terminateOnNoiseReached;
+	}
+
+	public boolean isReuseExistingDescription() {
+		return reuseExistingDescription;
+	}
+
+	public void setReuseExistingDescription(boolean reuseExistingDescription) {
+		this.reuseExistingDescription = reuseExistingDescription;
+	}
+
+	public boolean isUseMinimizer() {
+		return useMinimizer;
+	}
+
+	public void setUseMinimizer(boolean useMinimizer) {
+		this.useMinimizer = useMinimizer;
+	}
+
+	public OEHeuristicRuntime getHeuristic() {
+		return heuristic;
+	}
+
+	public void setHeuristic(OEHeuristicRuntime heuristic) {
+		this.heuristic = heuristic;
+	}	
+	
+	
+	public static void main(String[] args) throws Exception{
+		AbstractKnowledgeSource ks = new OWLFile("../examples/family/father_oe.owl");
+		ks.init();
+		
+		AbstractReasonerComponent rc = new FastInstanceChecker(ks);
+		rc.init();
+		
+		ClassLearningProblem lp = new ClassLearningProblem(rc);
+		lp.setClassToDescribe(new NamedClass("http://example.com/father#father"));
+		lp.init();
+		
+		PCELOE alg = new PCELOE(lp, rc);
+		alg.setMaxExecutionTimeInSeconds(10);
+		alg.setMaxClassDescriptionTests(200);
+		alg.init();
+		
+		alg.start();
+		
+	}
+	
+}

This was sent by the SourceForge.net collaborative development platform, the world's largest Open Source development site.





Thread: [DL-Learner SVN] SF.net SVN: dl-learner:[3529] trunk/components-core/src/main/java/org/ dllearner/

dl-learner-svn