You are viewing a plain text version of this content. The canonical link for it is here.
Posted to commits@opennlp.apache.org by jo...@apache.org on 2013/12/03 16:09:58 UTC
svn commit: r1547420 - in /opennlp/addons/liblinear-addon: ./
LiblinearParams.txt pom.xml src/ src/main/ src/main/java/
src/main/java/LiblinearModel.java src/main/java/LiblinearModelSerializer.java
src/main/java/LiblinearTrainer.java
Author: joern
Date: Tue Dec 3 15:09:57 2013
New Revision: 1547420
URL: http://svn.apache.org/r1547420
Log:
OPENNLP-624 Initial check in of the liblinear integration
Added:
opennlp/addons/liblinear-addon/
opennlp/addons/liblinear-addon/LiblinearParams.txt
opennlp/addons/liblinear-addon/pom.xml
opennlp/addons/liblinear-addon/src/
opennlp/addons/liblinear-addon/src/main/
opennlp/addons/liblinear-addon/src/main/java/
opennlp/addons/liblinear-addon/src/main/java/LiblinearModel.java
opennlp/addons/liblinear-addon/src/main/java/LiblinearModelSerializer.java
opennlp/addons/liblinear-addon/src/main/java/LiblinearTrainer.java
Added: opennlp/addons/liblinear-addon/LiblinearParams.txt
URL: http://svn.apache.org/viewvc/opennlp/addons/liblinear-addon/LiblinearParams.txt?rev=1547420&view=auto
==============================================================================
--- opennlp/addons/liblinear-addon/LiblinearParams.txt (added)
+++ opennlp/addons/liblinear-addon/LiblinearParams.txt Tue Dec 3 15:09:57 2013
@@ -0,0 +1,20 @@
+# Licensed to the Apache Software Foundation (ASF) under one or more
+# contributor license agreements. See the NOTICE file distributed with
+# this work for additional information regarding copyright ownership.
+# The ASF licenses this file to You under the Apache License, Version 2.0
+# (the "License"); you may not use this file except in compliance with
+# the License. You may obtain a copy of the License at
+#
+# http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+
+# Sample machine learning properties file
+
+Algorithm=LiblinearTrainer
+Iterations=100
+Cutoff=0
Added: opennlp/addons/liblinear-addon/pom.xml
URL: http://svn.apache.org/viewvc/opennlp/addons/liblinear-addon/pom.xml?rev=1547420&view=auto
==============================================================================
--- opennlp/addons/liblinear-addon/pom.xml (added)
+++ opennlp/addons/liblinear-addon/pom.xml Tue Dec 3 15:09:57 2013
@@ -0,0 +1,95 @@
+<?xml version="1.0" encoding="UTF-8"?>
+
+<!--
+ Licensed to the Apache Software Foundation (ASF) under one
+ or more contributor license agreements. See the NOTICE file
+ distributed with this work for additional information
+ regarding copyright ownership. The ASF licenses this file
+ to you under the Apache License, Version 2.0 (the
+ "License"); you may not use this file except in compliance
+ with the License. You may obtain a copy of the License at
+
+ http://www.apache.org/licenses/LICENSE-2.0
+
+ Unless required by applicable law or agreed to in writing,
+ software distributed under the License is distributed on an
+ "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY
+ KIND, either express or implied. See the License for the
+ specific language governing permissions and limitations
+ under the License.
+-->
+
+<project xmlns="http://maven.apache.org/POM/4.0.0" xmlns:xsi="http://www.w3.org/2001/XMLSchema-instance" xsi:schemaLocation="http://maven.apache.org/POM/4.0.0 http://maven.apache.org/xsd/maven-4.0.0.xsd">
+ <modelVersion>4.0.0</modelVersion>
+
+ <parent>
+ <groupId>org.apache.opennlp</groupId>
+ <artifactId>opennlp</artifactId>
+ <version>1.6.0-SNAPSHOT</version>
+ <relativePath>../opennlp/pom.xml</relativePath>
+ </parent>
+
+ <artifactId>opennlp-liblinear-addon</artifactId>
+ <packaging>jar</packaging>
+ <name>Apache OpenNLP Liblinear Addon</name>
+
+ <repositories>
+ <repository>
+ <id>ApacheIncubatorRepository</id>
+ <url>
+ http://people.apache.org/repo/m2-incubating-repository/
+ </url>
+ </repository>
+ </repositories>
+
+ <dependencies>
+ <dependency>
+ <groupId>org.apache.opennlp</groupId>
+ <artifactId>opennlp-tools</artifactId>
+ <version>1.6.0-SNAPSHOT</version>
+ </dependency>
+
+ <dependency>
+ <groupId>de.bwaldvogel</groupId>
+ <artifactId>liblinear</artifactId>
+ <version>1.92</version>
+ </dependency>
+
+ <dependency>
+ <groupId>junit</groupId>
+ <artifactId>junit</artifactId>
+ <scope>test</scope>
+ </dependency>
+ </dependencies>
+
+ <build>
+ <plugins>
+ <plugin>
+ <groupId>org.apache.maven.plugins</groupId>
+ <artifactId>maven-dependency-plugin</artifactId>
+ <version>2.1</version>
+ <executions>
+ <execution>
+ <id>copy-dependencies</id>
+ <phase>package</phase>
+ <goals>
+ <goal>copy-dependencies</goal>
+ </goals>
+ <configuration>
+ <excludeScope>provided</excludeScope>
+ <stripVersion>true</stripVersion>
+ </configuration>
+ </execution>
+ </executions>
+ </plugin>
+ <plugin>
+ <groupId>org.apache.maven.plugins</groupId>
+ <artifactId>maven-surefire-plugin</artifactId>
+ <configuration>
+ <skipTests>true</skipTests>
+ <argLine>-Xmx512m</argLine>
+ </configuration>
+ </plugin>
+ </plugins>
+ </build>
+</project>
Added: opennlp/addons/liblinear-addon/src/main/java/LiblinearModel.java
URL: http://svn.apache.org/viewvc/opennlp/addons/liblinear-addon/src/main/java/LiblinearModel.java?rev=1547420&view=auto
==============================================================================
--- opennlp/addons/liblinear-addon/src/main/java/LiblinearModel.java (added)
+++ opennlp/addons/liblinear-addon/src/main/java/LiblinearModel.java Tue Dec 3 15:09:57 2013
@@ -0,0 +1,143 @@
+/*
+ * Licensed to the Apache Software Foundation (ASF) under one
+ * or more contributor license agreements. See the NOTICE file
+ * distributed with this work for additional information
+ * regarding copyright ownership. The ASF licenses this file
+ * to you under the Apache License, Version 2.0 (the
+ * "License"); you may not use this file except in compliance
+ * with the License. You may obtain a copy of the License at
+ *
+ * http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing,
+ * software distributed under the License is distributed on an
+ * "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY
+ * KIND, either express or implied. See the License for the
+ * specific language governing permissions and limitations
+ * under the License.
+ */
+
+import java.io.IOException;
+import java.io.InputStream;
+import java.io.InputStreamReader;
+import java.io.OutputStream;
+import java.util.ArrayList;
+import java.util.List;
+import java.util.Map;
+
+import opennlp.tools.ml.model.MaxentModel;
+import opennlp.tools.util.model.ArtifactSerializer;
+import opennlp.tools.util.model.SerializableArtifact;
+import de.bwaldvogel.liblinear.Feature;
+import de.bwaldvogel.liblinear.FeatureNode;
+import de.bwaldvogel.liblinear.Linear;
+import de.bwaldvogel.liblinear.Model;
+
+// TODO: The features need to be serialized with the model
+// the liblinear model only contains the ints and weights,
+// but the string lables get lost ... basically that are two maps.
+
+// One for outcomes, one for the features ...
+
+public class LiblinearModel implements MaxentModel, SerializableArtifact {
+
+ private Model model;
+
+ // Lets read them from disk, when model is loaded ...
+ private String outcomeLabels[];
+ private Map<String, Integer> predMap;
+
+ public LiblinearModel(Model model, String outcomes[], Map<String, Integer> predMap) {
+ this.model = model;
+ this.outcomeLabels = outcomes;
+ this.predMap = predMap;
+ }
+
+ public LiblinearModel(InputStream in) throws IOException {
+ model = Linear.loadModel(new InputStreamReader(in));
+ }
+
+ public double[] eval(String[] features) {
+
+ // Note: If a feature can't be mapped, it will be ignored!
+
+ List<Integer> context = new ArrayList<Integer>(features.length);
+
+ for (int i = 0; i < features.length; i++) {
+ Integer feature = predMap.get(features[i]);
+
+ if (feature != null) {
+ context.add(feature);
+ }
+ }
+
+ return eval(context);
+ }
+
+ public double[] eval(String[] context, double[] probs) {
+ return eval(context);
+ }
+
+ public double[] eval(String[] context, float[] values) {
+ return eval(context);
+ }
+
+ private double[] eval(List<Integer> context) {
+
+ double outcomes[] = new double[outcomeLabels.length];
+
+ Feature vx[] = new Feature[context.size()];
+
+ for (int i = 0; i < context.size(); i++) {
+ vx[i] = new FeatureNode(context.get(i) + 1, 1d);
+ }
+
+ Linear.predictProbability(model, vx, outcomes);
+
+ return outcomes;
+ }
+
+ public String getAllOutcomes(double[] outcomes) {
+ // TODO: Return prev outcomes ..
+ return null;
+ }
+
+ public String getBestOutcome(double[] ocs) {
+ int best = 0;
+ for (int i = 1; i < ocs.length; i++)
+ if (ocs[i] > ocs[best]) best = i;
+ return outcomeLabels[best];
+ }
+
+ // TODO: This method needs to go away from the interface ... !!!
+ public Object[] getDataStructures() {
+ return null;
+ }
+
+ public int getIndex(String outcome) {
+ for (int i = 0; i < outcomeLabels.length; i++) {
+ if (outcomeLabels[i].equals(outcome)) {
+ return i;
+ }
+ }
+
+ return -1;
+ }
+
+ public int getNumOutcomes() {
+ return outcomeLabels.length;
+ }
+
+ public String getOutcome(int i) {
+ return outcomeLabels[i];
+ }
+
+ public void serialize(OutputStream out) throws IOException {
+
+ }
+
+ public Class<?> getSerializerClass() {
+ return LiblinearModelSerializer.class;
+ }
+
+}
Added: opennlp/addons/liblinear-addon/src/main/java/LiblinearModelSerializer.java
URL: http://svn.apache.org/viewvc/opennlp/addons/liblinear-addon/src/main/java/LiblinearModelSerializer.java?rev=1547420&view=auto
==============================================================================
--- opennlp/addons/liblinear-addon/src/main/java/LiblinearModelSerializer.java (added)
+++ opennlp/addons/liblinear-addon/src/main/java/LiblinearModelSerializer.java Tue Dec 3 15:09:57 2013
@@ -0,0 +1,39 @@
+/*
+ * Licensed to the Apache Software Foundation (ASF) under one
+ * or more contributor license agreements. See the NOTICE file
+ * distributed with this work for additional information
+ * regarding copyright ownership. The ASF licenses this file
+ * to you under the Apache License, Version 2.0 (the
+ * "License"); you may not use this file except in compliance
+ * with the License. You may obtain a copy of the License at
+ *
+ * http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing,
+ * software distributed under the License is distributed on an
+ * "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY
+ * KIND, either express or implied. See the License for the
+ * specific language governing permissions and limitations
+ * under the License.
+ */
+
+import java.io.IOException;
+import java.io.InputStream;
+import java.io.OutputStream;
+
+import opennlp.tools.util.InvalidFormatException;
+import opennlp.tools.util.model.ArtifactSerializer;
+
+public class LiblinearModelSerializer implements
+ ArtifactSerializer<LiblinearModel> {
+
+ public LiblinearModel create(InputStream in) throws IOException,
+ InvalidFormatException {
+ return new LiblinearModel(in);
+ }
+
+ public void serialize(LiblinearModel model, OutputStream out)
+ throws IOException {
+ model.serialize(out);
+ }
+}
Added: opennlp/addons/liblinear-addon/src/main/java/LiblinearTrainer.java
URL: http://svn.apache.org/viewvc/opennlp/addons/liblinear-addon/src/main/java/LiblinearTrainer.java?rev=1547420&view=auto
==============================================================================
--- opennlp/addons/liblinear-addon/src/main/java/LiblinearTrainer.java (added)
+++ opennlp/addons/liblinear-addon/src/main/java/LiblinearTrainer.java Tue Dec 3 15:09:57 2013
@@ -0,0 +1,180 @@
+/*
+ * Licensed to the Apache Software Foundation (ASF) under one
+ * or more contributor license agreements. See the NOTICE file
+ * distributed with this work for additional information
+ * regarding copyright ownership. The ASF licenses this file
+ * to you under the Apache License, Version 2.0 (the
+ * "License"); you may not use this file except in compliance
+ * with the License. You may obtain a copy of the License at
+ *
+ * http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing,
+ * software distributed under the License is distributed on an
+ * "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY
+ * KIND, either express or implied. See the License for the
+ * specific language governing permissions and limitations
+ * under the License.
+ */
+
+import java.io.BufferedWriter;
+import java.io.File;
+import java.io.FileWriter;
+import java.io.IOException;
+import java.util.ArrayList;
+import java.util.Collection;
+import java.util.HashMap;
+import java.util.List;
+import java.util.Map;
+
+import de.bwaldvogel.liblinear.Feature;
+import de.bwaldvogel.liblinear.FeatureNode;
+import de.bwaldvogel.liblinear.Linear;
+import de.bwaldvogel.liblinear.Model;
+import de.bwaldvogel.liblinear.Parameter;
+import de.bwaldvogel.liblinear.Problem;
+import de.bwaldvogel.liblinear.SolverType;
+import de.bwaldvogel.liblinear.Train;
+import opennlp.tools.ml.AbstractEventTrainer;
+import opennlp.tools.ml.model.DataIndexer;
+import opennlp.tools.ml.model.MaxentModel;
+
+public class LiblinearTrainer extends AbstractEventTrainer {
+
+ public LiblinearTrainer(Map<String, String> trainParams,
+ Map<String, String> reportMap) {
+ super(trainParams, reportMap);
+
+ // TODO: Extract solver type here
+ // depending on it, extract parameters
+ // e.g. bias, C, eps for L1_LR
+
+ }
+
+ private static Problem constructProblem(List<Double> vy, List<Feature[]> vx, int maxIndex, double bias) {
+
+ // Initialize problem
+ Problem problem = new Problem();
+ problem.l = vy.size();
+ problem.n = maxIndex;
+ problem.bias = bias;
+
+ if (bias >= 0) {
+ problem.n++;
+ }
+
+ problem.x = new Feature[problem.l][];
+
+ for (int i = 0; i < problem.l; i++) {
+ problem.x[i] = vx.get(i);
+
+ if (bias >= 0) {
+ problem.x[i][problem.x[i].length - 1] = new FeatureNode(max_index + 1, bias);
+ }
+ }
+
+ problem.y = new double[problem.l];
+
+ for (int i = 0; i < problem.l; i++) {
+ problem.y[i] = vy.get(i).doubleValue();
+ }
+
+ return problem;
+ }
+
+ @Override
+ public MaxentModel doTrain(DataIndexer indexer) throws IOException {
+
+ List<Double> vy = new ArrayList<Double>();
+ List<Feature[]> vx = new ArrayList<Feature[]>();
+
+ // outcomes
+ int outcomes[] = indexer.getOutcomeList();
+
+ final int bias = 0;
+
+ int max_index = 0;
+
+ // For each event ...
+ for (int i = 0; i < indexer.getContexts().length; i++) {
+
+ int outcome = outcomes[i];
+ vy.add(Double.valueOf(outcome));
+
+ int features[] = indexer.getContexts()[i];
+
+ Feature[] x;
+ if (bias >= 0) {
+ x = new Feature[features.length + 1];
+ } else {
+ x = new Feature[features.length];
+ }
+
+ // for each feature ...
+ for (int fi = 0; fi < features.length; fi++) {
+ x[fi] = new FeatureNode(features[fi] + 1, indexer.getNumTimesEventsSeen()[fi]);
+ }
+
+ if (features.length > 0) {
+ max_index = Math.max(max_index, x[features.length - 1].getIndex());
+ }
+
+ vx.add(x);
+ }
+
+ Problem problem = constructProblem(vy, vx, max_index, bias);
+ Parameter parameter = new Parameter(SolverType.L1R_LR, 1d, 0.001d);
+
+ Model liblinearModel = Linear.train(problem, parameter);
+
+ Map<String, Integer> predMap = new HashMap<String, Integer>();
+
+ String predLabels[] = indexer.getPredLabels();
+ for (int i = 0; i < predLabels.length; i++) {
+ predMap.put(predLabels[i], i);
+ }
+
+ return new LiblinearModel(liblinearModel, indexer.getOutcomeLabels(), predMap);
+ }
+
+ @Override
+ public boolean isSortAndMerge() {
+ return true;
+ }
+
+ public static void main(String[] args) throws Exception {
+
+ File file = File.createTempFile("svm", "test");
+ file.deleteOnExit();
+
+ Collection<String> lines = new ArrayList<String>();
+ lines.add("1 1:1 3:1 4:1 6:1");
+ lines.add("2 2:1 3:1 5:1 7:1");
+ lines.add("1 3:1 5:1");
+ lines.add("1 1:1 4:1 7:1");
+ lines.add("2 4:1 5:1 7:1");
+ lines.add("1 1:1 4:1 7:1");
+ lines.add("2 4:1 5:1 7:1");
+
+ BufferedWriter writer = new BufferedWriter(new FileWriter(file));
+ try {
+ for (String line : lines)
+ writer.append(line).append("\n");
+ } finally {
+ writer.close();
+ }
+
+ Train train = new Train();
+
+ Problem problem = train.readProblem(file, 0d);
+
+ Model model = Linear.train(problem, new Parameter(SolverType.L1R_LR, 10d,
+ 0.02d));
+
+ double result = Linear.predict(model, new Feature[]{new FeatureNode(4, 1d), new FeatureNode(1, 1d)});
+ double outcomes[] = new double[2];
+ double result2 = Linear.predictProbability(model, new Feature[]{new FeatureNode(4, 1d), new FeatureNode(1, 1d)}, outcomes);
+
+ System.out.println(result);
+ }
+}