Menu
▾
▴
[DL-Learner SVN] SF.net SVN: dl-learner:[1518] trunk
|
From: <jen...@us...> - 2008-11-16 19:20:38
|
Revision: 1518
http://dl-learner.svn.sourceforge.net/dl-learner/?rev=1518&view=rev
Author: jenslehmann
Date: 2008-11-16 19:20:34 +0000 (Sun, 16 Nov 2008)
Log Message:
-----------
- cardinality limit option
- reasoner unit test extended
- small fixes
Modified Paths:
--------------
trunk/examples/carcinogenesis/train.conf
trunk/src/dl-learner/org/dllearner/algorithms/refexamples/ExampleBasedNode.java
trunk/src/dl-learner/org/dllearner/algorithms/refexamples/ExampleBasedROLComponent.java
trunk/src/dl-learner/org/dllearner/algorithms/refexamples/ExampleBasedROLearner.java
trunk/src/dl-learner/org/dllearner/algorithms/refexamples/MultiHeuristic.java
trunk/src/dl-learner/org/dllearner/core/ReasonerComponent.java
trunk/src/dl-learner/org/dllearner/core/configurators/ExampleBasedROLComponentConfigurator.java
trunk/src/dl-learner/org/dllearner/core/options/CommonConfigOptions.java
trunk/src/dl-learner/org/dllearner/refinementoperators/RhoDRDown.java
trunk/src/dl-learner/org/dllearner/test/junit/ReasonerTests.java
Modified: trunk/examples/carcinogenesis/train.conf
===================================================================
--- trunk/examples/carcinogenesis/train.conf 2008-11-16 14:38:32 UTC (rev 1517)
+++ trunk/examples/carcinogenesis/train.conf 2008-11-16 19:20:34 UTC (rev 1518)
@@ -1,15 +1,21 @@
import("carcinogenesis.owl");
-// store some settings to make the experiment reproducable in the future
+// general setup
reasoner = fastInstanceChecker;
algorithm = refexamples;
refexamples.noisePercentage = 32;
refexamples.startClass = "http://dl-learner.org/carcinogenesis#Compound";
+
+// search tree
refexamples.writeSearchTree = false;
refexamples.searchTreeFile = "log/carcinogenesis/searchTree.log";
+
+// store some default settings to make the experiment reproducable in the future
refexamples.negativeWeight = 0.8;
-refexamples.startNodeBonus = 2.0;
+refexamples.startNodeBonus = 1.0;
refexamples.forceRefinementLengthIncrease = false;
+refexamples.cardinalityLimit = 5;
+refexamples.usePropernessChecks = false;
+"http://dl-learner.org/carcinogenesis#d1"
+"http://dl-learner.org/carcinogenesis#d10"
Modified: trunk/src/dl-learner/org/dllearner/algorithms/refexamples/ExampleBasedNode.java
===================================================================
--- trunk/src/dl-learner/org/dllearner/algorithms/refexamples/ExampleBasedNode.java 2008-11-16 14:38:32 UTC (rev 1517)
+++ trunk/src/dl-learner/org/dllearner/algorithms/refexamples/ExampleBasedNode.java 2008-11-16 19:20:34 UTC (rev 1518)
@@ -25,6 +25,7 @@
import java.util.SortedSet;
import java.util.TreeSet;
+import org.dllearner.core.configurators.ExampleBasedROLComponentConfigurator;
import org.dllearner.core.owl.Description;
import org.dllearner.core.owl.Individual;
import org.dllearner.utilities.owl.ConceptComparator;
@@ -41,6 +42,8 @@
*/
public class ExampleBasedNode {
+ private ExampleBasedROLComponentConfigurator configurator;
+
private static DecimalFormat df = new DecimalFormat();
// example based variables
@@ -72,7 +75,8 @@
// a flag whether this could be a solution for a posonly learning problem
private boolean isPosOnlyCandidate = true;
- public ExampleBasedNode(Description concept) {
+ public ExampleBasedNode(ExampleBasedROLComponentConfigurator configurator, Description concept) {
+ this.configurator = configurator;
this.concept = concept;
horizontalExpansion = 0;
isQualityEvaluated = false;
@@ -165,7 +169,7 @@
ret += "acc:" + df.format(accuracy) + "% ";
// comment this out to display the heuristic score with default parameters
- double heuristicScore = MultiHeuristic.getNodeScore(this, nrOfPositiveExamples, nrOfNegativeExamples);
+ double heuristicScore = MultiHeuristic.getNodeScore(this, nrOfPositiveExamples, nrOfNegativeExamples, configurator);
ret += "h:" +df.format(heuristicScore) + " ";
int wrongPositives = nrOfPositiveExamples - coveredPositives.size();
@@ -188,7 +192,7 @@
ret += "<b>acc: " + df.format(accuracy) + "% </b>";
// comment this out to display the heuristic score with default parameters
- double heuristicScore = MultiHeuristic.getNodeScore(this, nrOfPositiveExamples, nrOfNegativeExamples);
+ double heuristicScore = MultiHeuristic.getNodeScore(this, nrOfPositiveExamples, nrOfNegativeExamples, configurator);
ret += "h:" +df.format(heuristicScore) + " ";
int wrongPositives = nrOfPositiveExamples - coveredPositives.size();
@@ -212,7 +216,7 @@
ret += "acc:" + df.format(accuracy) + "% ";
// comment this out to display the heuristic score with default parameters
- double heuristicScore = MultiHeuristic.getNodeScore(this, nrOfPositiveExamples, nrOfNegativeExamples);
+ double heuristicScore = MultiHeuristic.getNodeScore(this, nrOfPositiveExamples, nrOfNegativeExamples, configurator);
ret += "h:" +df.format(heuristicScore) + " ";
int wrongPositives = nrOfPositiveExamples - coveredPositives.size();
Modified: trunk/src/dl-learner/org/dllearner/algorithms/refexamples/ExampleBasedROLComponent.java
===================================================================
--- trunk/src/dl-learner/org/dllearner/algorithms/refexamples/ExampleBasedROLComponent.java 2008-11-16 14:38:32 UTC (rev 1517)
+++ trunk/src/dl-learner/org/dllearner/algorithms/refexamples/ExampleBasedROLComponent.java 2008-11-16 19:20:34 UTC (rev 1518)
@@ -196,6 +196,7 @@
options.add(CommonConfigOptions.useHasValueConstructor());
options.add(CommonConfigOptions.valueFreqencyThreshold());
options.add(CommonConfigOptions.useCardinalityRestrictions());
+ options.add(CommonConfigOptions.cardinalityLimit());
options.add(CommonConfigOptions.useNegation());
options.add(CommonConfigOptions.useBooleanDatatypes());
options.add(CommonConfigOptions.useDoubleDatatypes());
@@ -376,6 +377,7 @@
RhoDRDown operator = new RhoDRDown(
reasoner,
classHierarchy,
+ configurator,
applyAllFilter,
applyExistsFilter,
useAllConstructor,
@@ -392,6 +394,7 @@
// create an algorithm object and pass all configuration
// options to it
algorithm = new ExampleBasedROLearner(
+ configurator,
learningProblem,
reasoner,
operator,
Modified: trunk/src/dl-learner/org/dllearner/algorithms/refexamples/ExampleBasedROLearner.java
===================================================================
--- trunk/src/dl-learner/org/dllearner/algorithms/refexamples/ExampleBasedROLearner.java 2008-11-16 14:38:32 UTC (rev 1517)
+++ trunk/src/dl-learner/org/dllearner/algorithms/refexamples/ExampleBasedROLearner.java 2008-11-16 19:20:34 UTC (rev 1518)
@@ -37,6 +37,7 @@
import org.dllearner.core.LearningProblem;
import org.dllearner.core.ReasonerComponent;
import org.dllearner.core.Score;
+import org.dllearner.core.configurators.ExampleBasedROLComponentConfigurator;
import org.dllearner.core.owl.Description;
import org.dllearner.core.owl.Individual;
import org.dllearner.core.owl.Intersection;
@@ -73,6 +74,7 @@
public class ExampleBasedROLearner {
private static Logger logger = Logger.getLogger(ExampleBasedROLearner.class);
+ private ExampleBasedROLComponentConfigurator configurator;
// basic setup: learning problem and reasoning service
private ReasonerComponent rs;
@@ -232,6 +234,7 @@
private Map<String, String> prefixes;
public ExampleBasedROLearner(
+ ExampleBasedROLComponentConfigurator configurator,
LearningProblem learningProblem,
ReasonerComponent rs,
RefinementOperator operator,
@@ -268,6 +271,7 @@
// nrOfNegativeExamples = lp.getPseudoNegatives().size();
nrOfNegativeExamples = 0;
}
+ this.configurator = configurator;
nrOfExamples = nrOfPositiveExamples + nrOfNegativeExamples;
this.rs = rs;
this.operator = (RhoDRDown) operator;
@@ -384,10 +388,10 @@
// start search with start class
if (startDescription == null) {
- startNode = new ExampleBasedNode(Thing.instance);
+ startNode = new ExampleBasedNode(configurator, Thing.instance);
startNode.setCoveredExamples(positiveExamples, negativeExamples);
} else {
- startNode = new ExampleBasedNode(startDescription);
+ startNode = new ExampleBasedNode(configurator, startDescription);
Set<Individual> coveredNegatives = rs.hasType(startDescription, negativeExamples);
Set<Individual> coveredPositives = rs.hasType(startDescription, positiveExamples);
startNode.setCoveredExamples(coveredPositives, coveredNegatives);
@@ -673,7 +677,7 @@
properRefinements.add(refinement);
tooWeakList.add(refinement);
- ExampleBasedNode newNode = new ExampleBasedNode(refinement);
+ ExampleBasedNode newNode = new ExampleBasedNode(configurator, refinement);
newNode.setHorizontalExpansion(refinement.getLength() - 1);
newNode.setTooWeak(true);
newNode
@@ -763,7 +767,7 @@
if (nonRedundant) {
// newly created node
- ExampleBasedNode newNode = new ExampleBasedNode(refinement);
+ ExampleBasedNode newNode = new ExampleBasedNode(configurator, refinement);
// die -1 ist wichtig, da sonst keine gleich langen Refinements
// für den neuen Knoten erlaubt wären z.B. person => male
newNode.setHorizontalExpansion(refinement.getLength() - 1);
Modified: trunk/src/dl-learner/org/dllearner/algorithms/refexamples/MultiHeuristic.java
===================================================================
--- trunk/src/dl-learner/org/dllearner/algorithms/refexamples/MultiHeuristic.java 2008-11-16 14:38:32 UTC (rev 1517)
+++ trunk/src/dl-learner/org/dllearner/algorithms/refexamples/MultiHeuristic.java 2008-11-16 19:20:34 UTC (rev 1518)
@@ -69,6 +69,7 @@
public class MultiHeuristic implements ExampleBasedHeuristic {
private ConceptComparator conceptComparator = new ConceptComparator();
+// private ExampleBasedROLComponentConfigurator configurator;
// heuristic parameters
private double expansionPenaltyFactor = 0.02;
@@ -83,6 +84,7 @@
private int nrOfNegativeExamples;
private int nrOfExamples;
+ @Deprecated
public MultiHeuristic(int nrOfPositiveExamples, int nrOfNegativeExamples) {
this.nrOfNegativeExamples = nrOfNegativeExamples;
nrOfExamples = nrOfPositiveExamples + nrOfNegativeExamples;
@@ -92,10 +94,9 @@
public MultiHeuristic(int nrOfPositiveExamples, int nrOfNegativeExamples, ExampleBasedROLComponentConfigurator configurator) {
this.nrOfNegativeExamples = nrOfNegativeExamples;
nrOfExamples = nrOfPositiveExamples + nrOfNegativeExamples;
+// this.configurator = configurator;
negativeWeight = configurator.getNegativeWeight();
startNodeBonus = configurator.getStartNodeBonus();
- System.out.println(negativeWeight);
- System.out.println(startNodeBonus);
}
// public MultiHeuristic(int nrOfPositiveExamples, int nrOfNegativeExamples, double expansionPenaltyFactor, double gainBonusFactor) {
@@ -141,8 +142,8 @@
return (coveredPositives + negativeWeight * (nrOfNegativeExamples - coveredNegatives))/(double)nrOfExamples;
}
- public static double getNodeScore(ExampleBasedNode node, int nrOfPositiveExamples, int nrOfNegativeExamples) {
- MultiHeuristic multi = new MultiHeuristic(nrOfPositiveExamples, nrOfNegativeExamples);
+ public static double getNodeScore(ExampleBasedNode node, int nrOfPositiveExamples, int nrOfNegativeExamples, ExampleBasedROLComponentConfigurator configurator) {
+ MultiHeuristic multi = new MultiHeuristic(nrOfPositiveExamples, nrOfNegativeExamples, configurator);
return multi.getNodeScore(node);
}
Modified: trunk/src/dl-learner/org/dllearner/core/ReasonerComponent.java
===================================================================
--- trunk/src/dl-learner/org/dllearner/core/ReasonerComponent.java 2008-11-16 14:38:32 UTC (rev 1517)
+++ trunk/src/dl-learner/org/dllearner/core/ReasonerComponent.java 2008-11-16 19:20:34 UTC (rev 1518)
@@ -601,8 +601,7 @@
SortedSet<Constant> values = e.getValue();
if (values.size() > 1) {
logger.warn("Property " + datatypeProperty + " has more than one value " + e.getValue()
- + " for individual " + e.getKey() + ". We ignore the value.");
- // d135
+ + " for individual " + e.getKey() + ". We ignore the value.");
} else {
if (values.first().getLiteral().equalsIgnoreCase("true")) {
ret.add(e.getKey());
Modified: trunk/src/dl-learner/org/dllearner/core/configurators/ExampleBasedROLComponentConfigurator.java
===================================================================
--- trunk/src/dl-learner/org/dllearner/core/configurators/ExampleBasedROLComponentConfigurator.java 2008-11-16 14:38:32 UTC (rev 1517)
+++ trunk/src/dl-learner/org/dllearner/core/configurators/ExampleBasedROLComponentConfigurator.java 2008-11-16 19:20:34 UTC (rev 1518)
@@ -241,6 +241,15 @@
return (Boolean) ComponentManager.getInstance().getConfigOptionValue(exampleBasedROLComponent, "useCardinalityRestrictions") ;
}
/**
+* cardinalityLimit Gives the maximum number used in cardinality restrictions..
+* mandatory: false| reinit necessary: true
+* default value: 5
+* @return int
+**/
+public int getCardinalityLimit() {
+return (Integer) ComponentManager.getInstance().getConfigOptionValue(exampleBasedROLComponent, "cardinalityLimit") ;
+}
+/**
* useNegation specifies whether negation is used in the learning algorothm.
* mandatory: false| reinit necessary: true
* default value: true
@@ -557,6 +566,15 @@
reinitNecessary = true;
}
/**
+* @param cardinalityLimit Gives the maximum number used in cardinality restrictions..
+* mandatory: false| reinit necessary: true
+* default value: 5
+**/
+public void setCardinalityLimit(int cardinalityLimit) {
+ComponentManager.getInstance().applyConfigEntry(exampleBasedROLComponent, "cardinalityLimit", cardinalityLimit);
+reinitNecessary = true;
+}
+/**
* @param useNegation specifies whether negation is used in the learning algorothm.
* mandatory: false| reinit necessary: true
* default value: true
Modified: trunk/src/dl-learner/org/dllearner/core/options/CommonConfigOptions.java
===================================================================
--- trunk/src/dl-learner/org/dllearner/core/options/CommonConfigOptions.java 2008-11-16 14:38:32 UTC (rev 1517)
+++ trunk/src/dl-learner/org/dllearner/core/options/CommonConfigOptions.java 2008-11-16 19:20:34 UTC (rev 1518)
@@ -43,6 +43,7 @@
public static boolean useHasValueConstructorDefault = false;
public static int valueFrequencyThresholdDefault = 3;
public static boolean useCardinalityRestrictionsDefault = true;
+ public static int cardinalityLimitDefault = 5;
public static boolean useNegationDefault = true;
public static boolean useBooleanDatatypesDefault = true;
public static boolean useDoubleDatatypesDefault = true;
@@ -115,6 +116,10 @@
return new BooleanConfigOption("useCardinalityRestrictions", "specifies whether CardinalityRestrictions is used in the learning algorithm",useCardinalityRestrictionsDefault);
}
+ public static IntegerConfigOption cardinalityLimit() {
+ return new IntegerConfigOption("cardinalityLimit", "Gives the maximum number used in cardinality restrictions.",cardinalityLimitDefault);
+ }
+
public static BooleanConfigOption useNegation() {
return new BooleanConfigOption("useNegation", "specifies whether negation is used in the learning algorothm",useNegationDefault);
}
Modified: trunk/src/dl-learner/org/dllearner/refinementoperators/RhoDRDown.java
===================================================================
--- trunk/src/dl-learner/org/dllearner/refinementoperators/RhoDRDown.java 2008-11-16 14:38:32 UTC (rev 1517)
+++ trunk/src/dl-learner/org/dllearner/refinementoperators/RhoDRDown.java 2008-11-16 19:20:34 UTC (rev 1518)
@@ -35,6 +35,7 @@
import org.apache.log4j.Logger;
import org.dllearner.core.ReasonerComponent;
+import org.dllearner.core.configurators.ExampleBasedROLComponentConfigurator;
import org.dllearner.core.options.CommonConfigOptions;
import org.dllearner.core.owl.BooleanValueRestriction;
import org.dllearner.core.owl.DataRange;
@@ -102,8 +103,7 @@
// limit for cardinality restrictions (this makes sense if we e.g. have compounds with up to
// more than 200 atoms but we are only interested in atoms with certain characteristics and do
// not want something like e.g. >= 204 hasAtom.NOT Carbon-87; which blows up the search space
- //RBC
- private int cardinalityLimit = 2;
+ private int cardinalityLimit = 5;
// start concept (can be used to start from an arbitrary concept, needs
// to be Thing or NamedClass), note that when you use e.g. Compound as
@@ -182,14 +182,15 @@
// private Map<NamedClass,Map<NamedClass,Boolean>> notABMeaningful = new TreeMap<NamedClass,Map<NamedClass,Boolean>>();
public RhoDRDown(ReasonerComponent reasoningService) {
- this(reasoningService, reasoningService.getClassHierarchy(), true, true, true, true, true, 3, true, true, true, true, null);
+// this(reasoningService, reasoningService.getClassHierarchy(), null, true, true, true, true, true, 3, true, true, true, true, null);
+ this.rs = reasoningService;
}
// TODO constructor which takes a RhoDRDownConfigurator object;
// this should be an interface implemented e.g. by ExampleBasedROLComponentConfigurator;
// the goal is to use the configurator system while still being flexible enough to
// use one refinement operator in several learning algorithms
- public RhoDRDown(ReasonerComponent reasoningService, ClassHierarchy subHierarchy, boolean applyAllFilter, boolean applyExistsFilter, boolean useAllConstructor,
+ public RhoDRDown(ReasonerComponent reasoningService, ClassHierarchy subHierarchy, ExampleBasedROLComponentConfigurator configurator, boolean applyAllFilter, boolean applyExistsFilter, boolean useAllConstructor,
boolean useExistsConstructor, boolean useHasValueConstructor, int valueFrequencyThreshold, boolean useCardinalityRestrictions,boolean useNegation, boolean useBooleanDatatypes, boolean useDoubleDatatypes, NamedClass startClass) {
this.rs = reasoningService;
this.subHierarchy = subHierarchy;
@@ -200,6 +201,7 @@
this.useHasValueConstructor = useHasValueConstructor;
this.frequencyThreshold = valueFrequencyThreshold;
this.useCardinalityRestrictions = useCardinalityRestrictions;
+ cardinalityLimit = configurator.getCardinalityLimit();
this.useNegation = useNegation;
this.useBooleanDatatypes = useBooleanDatatypes;
this.useDoubleDatatypes = useDoubleDatatypes;
Modified: trunk/src/dl-learner/org/dllearner/test/junit/ReasonerTests.java
===================================================================
--- trunk/src/dl-learner/org/dllearner/test/junit/ReasonerTests.java 2008-11-16 14:38:32 UTC (rev 1517)
+++ trunk/src/dl-learner/org/dllearner/test/junit/ReasonerTests.java 2008-11-16 19:20:34 UTC (rev 1518)
@@ -21,6 +21,9 @@
import static org.junit.Assert.assertTrue;
+import java.io.File;
+import java.net.MalformedURLException;
+import java.util.LinkedList;
import java.util.List;
import org.apache.log4j.Logger;
@@ -32,8 +35,10 @@
import org.dllearner.core.owl.Individual;
import org.dllearner.core.owl.KB;
import org.dllearner.kb.KBFile;
+import org.dllearner.kb.OWLFile;
import org.dllearner.parser.KBParser;
import org.dllearner.parser.ParseException;
+import org.dllearner.reasoning.OWLAPIReasoner;
import org.junit.Test;
/**
@@ -45,6 +50,8 @@
public class ReasonerTests {
private static Logger logger = Logger.getLogger(ReasonerTests.class);
+
+ private String baseURI;
public KB getSimpleKnowledgeBase() {
String kb = "person SUB TOP.";
@@ -101,5 +108,65 @@
e.printStackTrace();
}
}
+
+ /**
+ * Test of fast instance check algorithm on carcinogenesis data set.
+ * @throws ComponentInitException
+ * @throws ParseException
+ */
+ @Test
+ public void fastInstanceCheckTest() throws ComponentInitException, ParseException {
+ String file = "examples/carcinogenesis/carcinogenesis.owl";
+ ComponentManager cm = ComponentManager.getInstance();
+ KnowledgeSource ks = cm.knowledgeSource(OWLFile.class);
+ try {
+ cm.applyConfigEntry(ks, "url", new File(file).toURI().toURL());
+ } catch (MalformedURLException e) {
+ // should never happen
+ e.printStackTrace();
+ }
+ ks.init();
+ ReasonerComponent reasoner = cm.reasoner(OWLAPIReasoner.class, ks);
+ reasoner.init();
+ baseURI = reasoner.getBaseURI();
+
+ List<Description> testDescriptions = new LinkedList<Description>();
+ List<List<Individual>> posIndividuals = new LinkedList<List<Individual>>();
+ List<List<Individual>> negIndividuals = new LinkedList<List<Individual>>();
+
+ // TODO manually verify that the results are indeed correct
+ testDescriptions.add(KBParser.parseConcept("\"http://dl-learner.org/carcinogenesis#Compound\" AND (\"http://dl-learner.org/carcinogenesis#amesTestPositive\" = true OR >= 2 \"http://dl-learner.org/carcinogenesis#hasStructure\" \"http://dl-learner.org/carcinogenesis#Ar_halide\"))"));
+ posIndividuals.add(getIndSet("d113","d133","d171","d262","d265","d294","d68","d77","d79"));
+ negIndividuals.add(getIndSet("d139","d199","d202","d203","d283","d42"));
+ // TODO add more descriptions and instances
+
+ // make the specified assertions
+ for(int i=0; i<testDescriptions.size(); i++) {
+ Description description = testDescriptions.get(i);
+ List<Individual> pos = posIndividuals.get(i);
+ List<Individual> neg = negIndividuals.get(i);
+
+ for(Individual ind : pos) {
+ assertTrue(reasoner.hasType(description, ind));
+ }
+
+ for(Individual ind : neg) {
+ assertTrue(!reasoner.hasType(description, ind));
+ }
+ }
+ }
+
+ private List<Individual> getIndSet(String... inds) {
+ List<Individual> individuals = new LinkedList<Individual>();
+ for(String ind : inds) {
+ individuals.add(new Individual(uri(ind)));
+ }
+ return individuals;
+ }
+
+ private String uri(String name) {
+ return "\""+baseURI+name+"\"";
+ }
+
}
This was sent by the SourceForge.net collaborative development platform, the world's largest Open Source development site.
|
