package org.maltparser.ml.libsvm;
import java.io.BufferedReader;
import java.io.BufferedWriter;
import java.io.File;
import java.io.FileNotFoundException;
import java.io.IOException;
import java.io.InputStream;
import java.io.InputStreamReader;
import java.io.OutputStreamWriter;
import java.io.PrintStream;
import java.text.DecimalFormat;
import java.text.DecimalFormatSymbols;
import java.util.ArrayList;
import java.util.HashMap;
import java.util.Map;
import java.util.Set;
import java.util.jar.JarEntry;
import java.util.regex.Pattern;
import java.util.regex.PatternSyntaxException;
import libsvm.svm;
import libsvm.svm_model;
import libsvm.svm_node;
import libsvm.svm_parameter;
import libsvm.svm_problem;
import org.maltparser.core.exception.MaltChainedException;
import org.maltparser.core.feature.FeatureVector;
import org.maltparser.core.feature.function.FeatureFunction;
import org.maltparser.core.feature.value.FeatureValue;
import org.maltparser.core.feature.value.MultipleFeatureValue;
import org.maltparser.core.feature.value.SingleFeatureValue;
import org.maltparser.core.helper.NoPrintStream;
import org.maltparser.core.syntaxgraph.DependencyStructure;
import org.maltparser.ml.LearningMethod;
import org.maltparser.ml.liblinear.LiblinearException;
import org.maltparser.ml.libsvm.LibsvmException;
import org.maltparser.parser.DependencyParserConfig;
import org.maltparser.parser.guide.instance.InstanceModel;
import org.maltparser.parser.history.action.SingleDecision;
import org.maltparser.parser.history.kbest.KBestList;
import org.maltparser.parser.history.kbest.ScoredKBestList;
/**
Implements an interface to the LIBSVM learner (currently the LIBSVM 2.91 is used). More information
about LIBSVM can be found at
<a href="http://www.csie.ntu.edu.tw/~cjlin/libsvm/" target="_blank">LIBSVM -- A Library for Support Vector Machines</a>.
@author Johan Hall
@since 1.0
*/
public class Libsvm implements LearningMethod {
public final static String LIBSVM_VERSION = "2.91";
public enum Verbostity {
SILENT, ERROR, ALL
}
protected InstanceModel owner;
protected int learnerMode;
protected String name;
protected int numberOfInstances;
protected boolean saveInstanceFiles;
protected boolean excludeNullValues;
protected String pathExternalSVMTrain = null;
private int[] cardinalities;
/**
* Instance output stream writer
*/
private BufferedWriter instanceOutput = null;
/**
* LIBSVM svm_model object, only used during classification.
*/
private svm_model model = null;
/**
* LIBSVM svm_parameter object
*/
private svm_parameter svmParam;
/**
* Parameter string
*/
private String paramString;
/**
* An array of LIBSVM svm_node objects, only used during classification.
*/
private ArrayList<svm_node> xlist = null;
private Verbostity verbosity;
/**
* Constructs a LIBSVM learner.
*
* @param owner the guide model owner
* @param learnerMode the mode of the learner TRAIN or CLASSIFY
*/
public Libsvm(InstanceModel owner, Integer learnerMode) throws MaltChainedException {
setOwner(owner);
setLearningMethodName("libsvm");
setLearnerMode(learnerMode.intValue());
setNumberOfInstances(0);
verbosity = Verbostity.SILENT;
initSvmParam(getConfiguration().getOptionValue("libsvm", "libsvm_options").toString());
initSpecialParameters();
if (learnerMode == BATCH) {
// if (owner.getGuide().getConfiguration().getConfigLogger().isInfoEnabled()) {
// if (pathExternalSVMTrain != null) {
// owner.getGuide().getConfiguration().getConfigLogger().info(" Learner : LIBSVM external "+ getParamString() + "\n");
// } else {
// owner.getGuide().getConfiguration().getConfigLogger().info(" Learner : LIBSVM "+LIBSVM_VERSION+" "+ getParamString() + "\n");
// }
// }
instanceOutput = new BufferedWriter(getInstanceOutputStreamWriter(".ins"));
}
// else {
// if (owner.getGuide().getConfiguration().getConfigLogger().isInfoEnabled()) {
// owner.getGuide().getConfiguration().getConfigLogger().info(" Classifier : LIBSVM "+LIBSVM_VERSION+" "+ getParamString()+ "\n");
// }
// }
}
public void addInstance(SingleDecision decision, FeatureVector featureVector) throws MaltChainedException {
if (featureVector == null) {
throw new LibsvmException("The feature vector cannot be found");
} else if (decision == null) {
throw new LibsvmException("The decision cannot be found");
}
try {
instanceOutput.write(decision.getDecisionCode()+"\t");
for (int i = 0; i < featureVector.size(); i++) {
FeatureValue featureValue = featureVector.get(i).getFeatureValue();
if (excludeNullValues == true && featureValue.isNullValue()) {
instanceOutput.write("-1");
} else {
if (featureValue instanceof SingleFeatureValue) {
instanceOutput.write(((SingleFeatureValue)featureValue).getCode()+"");
} else if (featureValue instanceof MultipleFeatureValue) {
Set<Integer> values = ((MultipleFeatureValue)featureValue).getCodes();
int j=0;
for (Integer value : values) {
instanceOutput.write(value.toString());
if (j != values.size()-1) {
instanceOutput.write("|");
}
j++;
}
}
}
if (i != featureVector.size()) {
instanceOutput.write('\t');
}
}
instanceOutput.write('\n');
instanceOutput.flush();
increaseNumberOfInstances();
} catch (IOException e) {
throw new LibsvmException("The LIBSVM learner cannot write to the instance file. ", e);
}
}
public void finalizeSentence(DependencyStructure dependencyGraph) throws MaltChainedException { }
/* (non-Javadoc)
* @see org.maltparser.ml.LearningMethod#noMoreInstances()
*/
public void noMoreInstances() throws MaltChainedException {
closeInstanceWriter();
}
/* (non-Javadoc)
* @see org.maltparser.ml.LearningMethod#train(org.maltparser.parser.guide.feature.FeatureVector)
*/
public void train(FeatureVector featureVector) throws MaltChainedException {
if (featureVector == null) {
throw new LibsvmException("The feature vector cannot be found. ");
} else if (owner == null) {
throw new LibsvmException("The parent guide model cannot be found. ");
}
cardinalities = getCardinalities(featureVector);
if (pathExternalSVMTrain == null) {
try {
final svm_problem prob = readProblemMaltSVMFormat(getInstanceInputStreamReader(".ins"), cardinalities, svmParam);
if(svm.svm_check_parameter(prob, svmParam) != null) {
throw new LibsvmException(svm.svm_check_parameter(prob, svmParam));
}
owner.getGuide().getConfiguration().getConfigLogger().info("Creating LIBSVM model "+getFile(".mod").getName()+"\n");
final PrintStream out = System.out;
final PrintStream err = System.err;
System.setOut(NoPrintStream.NO_PRINTSTREAM);
System.setErr(NoPrintStream.NO_PRINTSTREAM);
svm.svm_save_model(getFile(".mod").getAbsolutePath(), svm.svm_train(prob, svmParam));
System.setOut(err);
System.setOut(out);
if (!saveInstanceFiles) {
getFile(".ins").delete();
}
} catch (OutOfMemoryError e) {
throw new LibsvmException("Out of memory. Please increase the Java heap size (-Xmx<size>). ", e);
} catch (IllegalArgumentException e) {
throw new LibsvmException("The LIBSVM learner was not able to redirect Standard Error stream. ", e);
} catch (SecurityException e) {
throw new LibsvmException("The LIBSVM learner cannot remove the instance file. ", e);
} catch (IOException e) {
throw new LibsvmException("The LIBSVM learner cannot save the model file '"+getFile(".mod").getAbsolutePath()+"'. ", e);
}
} else {
trainExternal(featureVector);
}
saveCardinalities(getInstanceOutputStreamWriter(".car"), cardinalities);
}
@Override
public double crossValidate(FeatureVector featureVector, int nrOfSplits)
throws MaltChainedException {
if (featureVector == null) {
throw new LibsvmException("The feature vector cannot be found. ");
} else if (owner == null) {
throw new LibsvmException("The parent guide model cannot be found. ");
}
cardinalities = getCardinalities(featureVector);
//TODO Implement support for externial SVM for cross validation
//if (pathExternalSVMTrain == null) {
double crossValidationAccuracy = 0.0;
try {
final svm_problem prob = readProblemMaltSVMFormat(getInstanceInputStreamReader(".ins"), cardinalities, svmParam);
if(svm.svm_check_parameter(prob, svmParam) != null) {
throw new LibsvmException(svm.svm_check_parameter(prob, svmParam));
}
owner.getGuide().getConfiguration().getConfigLogger().info("Doing cross validation\n");
final PrintStream out = System.out;
final PrintStream err = System.err;
System.setOut(NoPrintStream.NO_PRINTSTREAM);
System.setErr(NoPrintStream.NO_PRINTSTREAM);
//svm.svm_save_model(getFile(".mod").getAbsolutePath(), svm.svm_train(prob, svmParam));
double[] target = new double[prob.l];
svm.svm_cross_validation(prob, svmParam, nrOfSplits, target);
System.setOut(err);
System.setOut(out);
if (!saveInstanceFiles) {
getFile(".ins").delete();
}
double total_correct = 0.0;
for(int i=0;i<prob.l;i++)
if(target[i] == prob.y[i])
++total_correct;
if(total_correct>0)
crossValidationAccuracy = 100.0*total_correct/prob.l;
} catch (OutOfMemoryError e) {
throw new LibsvmException("Out of memory. Please increase the Java heap size (-Xmx<size>). ", e);
} catch (IllegalArgumentException e) {
throw new LibsvmException("The LIBSVM learner was not able to redirect Standard Error stream. ", e);
} catch (SecurityException e) {
throw new LibsvmException("The LIBSVM learner cannot remove the instance file. ", e);
}
//} else {
// trainExternal(featureVector);
//}
//saveCardinalities(getInstanceOutputStreamWriter(".car"), cardinalities);
return crossValidationAccuracy;
}
private void trainExternal(FeatureVector featureVector) throws MaltChainedException {
try {
maltSVMFormat2OriginalSVMFormat(getInstanceInputStreamReader(".ins"), getInstanceOutputStreamWriter(".ins.tmp"), cardinalities);
owner.getGuide().getConfiguration().getConfigLogger().info("Creating LIBSVM model (svm-train) "+getFile(".mod").getName());
final ArrayList<String> commands = new ArrayList<String>();
commands.add(pathExternalSVMTrain);
final String[] params = getSVMParamStringArray(svmParam);
for (int i=0; i < params.length; i++) {
commands.add(params[i]);
}
commands.add(getFile(".ins.tmp").getAbsolutePath());
commands.add(getFile(".mod").getAbsolutePath());
String[] arrayCommands = commands.toArray(new String[commands.size()]);
if (verbosity == Verbostity.ALL) {
owner.getGuide().getConfiguration().getConfigLogger().info('\n');
}
final Process child = Runtime.getRuntime().exec(arrayCommands);
final InputStream in = child.getInputStream();
final InputStream err = child.getErrorStream();
int c;
while ((c = in.read()) != -1){
if (verbosity == Verbostity.ALL) {
owner.getGuide().getConfiguration().getConfigLogger().info((char)c);
}
}
while ((c = err.read()) != -1){
if (verbosity == Verbostity.ALL || verbosity == Verbostity.ERROR) {
owner.getGuide().getConfiguration().getConfigLogger().info((char)c);
}
}
if (child.waitFor() != 0) {
owner.getGuide().getConfiguration().getConfigLogger().info(" FAILED ("+child.exitValue()+")");
}
in.close();
err.close();
if (!saveInstanceFiles) {
getFile(".ins").delete();
getFile(".ins.tmp").delete();
}
owner.getGuide().getConfiguration().getConfigLogger().info('\n');
} catch (InterruptedException e) {
throw new LibsvmException("SVM-trainer is interrupted. ", e);
} catch (IllegalArgumentException e) {
throw new LibsvmException("The LIBSVM learner was not able to redirect Standard Error stream. ", e);
} catch (SecurityException e) {
throw new LibsvmException("The LIBSVM learner cannot remove the instance file. ", e);
} catch (IOException e) {
throw new LibsvmException("The LIBSVM learner cannot save the model file '"+getFile(".mod").getAbsolutePath()+"'. ", e);
} catch (OutOfMemoryError e) {
throw new LibsvmException("Out of memory. Please increase the Java heap size (-Xmx<size>). ", e);
}
}
private int[] getCardinalities(FeatureVector featureVector) {
int[] cardinalities = new int[featureVector.size()];
int i = 0;
for (FeatureFunction feature : featureVector) {
cardinalities[i++] = feature.getFeatureValue().getCardinality();
}
return cardinalities;
}
private void saveCardinalities(OutputStreamWriter osw, int[] cardinalities) throws MaltChainedException {
final BufferedWriter out = new BufferedWriter(osw);
try {
for (int i = 0, n = cardinalities.length; i < n; i++) {
out.write(Integer.toString(cardinalities[i]));
if (i < n - 1) {
out.write(',');
}
}
out.write('\n');
out.close();
} catch (IOException e) {
throw new LibsvmException("Couldn't save the cardinalities to file. ", e);
}
}
private int[] loadCardinalities(InputStreamReader isr) throws MaltChainedException {
int[] cardinalities = null;
try {
final BufferedReader in = new BufferedReader(isr);
String line;
if ((line = in.readLine()) != null) {
String[] items = line.split(",");
cardinalities = new int[items.length];
for (int i = 0; i < items.length; i++) {
cardinalities[i] = Integer.parseInt(items[i]);
}
}
in.close();
} catch (IOException e) {
throw new LibsvmException("The cardinalities cannot be read because wrongly formatted. ", e);
} catch (NumberFormatException e) {
throw new LibsvmException("Couldn't load the cardinalities from file. ", e);
}
return cardinalities;
}
/* (non-Javadoc)
* @see org.maltparser.ml.LearningMethod#moveAllInstances(org.maltparser.ml.LearningMethod, org.maltparser.core.feature.function.FeatureFunction, java.util.ArrayList)
*/
public void moveAllInstances(LearningMethod method, FeatureFunction divideFeature, ArrayList<Integer> divideFeatureIndexVector) throws MaltChainedException {
if (method == null) {
throw new LibsvmException("The learning method cannot be found. ");
} else if (divideFeature == null) {
throw new LibsvmException("The divide feature cannot be found. ");
}
try {
final BufferedReader in = new BufferedReader(getInstanceInputStreamReader(".ins"));
final BufferedWriter out = method.getInstanceWriter();
final StringBuilder sb = new StringBuilder(6);
int l = in.read();
char c;
int j = 0;
while(true) {
if (l == -1) {
sb.setLength(0);
break;
}
c = (char)l;
l = in.read();
if (c == '\t') {
if (divideFeatureIndexVector.contains(j-1)) {
out.write(Integer.toString(((SingleFeatureValue)divideFeature.getFeatureValue()).getCode()));
out.write('\t');
}
out.write(sb.toString());
j++;
out.write('\t');
sb.setLength(0);
} else if (c == '\n') {
if (sb.length() > 0) {
out.write(sb.toString());
}
if (divideFeatureIndexVector.contains(j-1)) {
if (sb.length() > 0) {
out.write('\t');
}
out.write(Integer.toString(((SingleFeatureValue)divideFeature.getFeatureValue()).getCode()));
}
out.write('\n');
sb.setLength(0);
method.increaseNumberOfInstances();
this.decreaseNumberOfInstances();
j = 0;
} else {
sb.append(c);
}
}
in.close();
getFile(".ins").delete();
} catch (SecurityException e) {
throw new LibsvmException("The LIBSVM learner cannot remove the instance file. ", e);
} catch (NullPointerException e) {
throw new LibsvmException("The instance file cannot be found. ", e);
} catch (FileNotFoundException e) {
throw new LibsvmException("The instance file cannot be found. ", e);
} catch (IOException e) {
throw new LibsvmException("The LIBSVM learner read from the instance file. ", e);
}
}
/* (non-Javadoc)
* @see org.maltparser.ml.LearningMethod#predict(org.maltparser.parser.guide.feature.FeatureVector, org.maltparser.ml.KBestList)
*/
public boolean predict(FeatureVector featureVector, SingleDecision decision) throws MaltChainedException {
if (model == null) {
try {
model = svm.svm_load_model(new BufferedReader(getInstanceInputStreamReaderFromConfigFile(".mod")));
} catch (IOException e) {
throw new LibsvmException("The model cannot be loaded. ", e);
}
}
if (cardinalities == null) {
if (getConfigFileEntry(".car") != null) {
cardinalities = loadCardinalities(getInstanceInputStreamReaderFromConfigFile(".car"));
} else {
cardinalities = getCardinalities(featureVector);
}
}
if (xlist == null) {
xlist = new ArrayList<svm_node>(featureVector.size());
}
if (model == null) {
throw new LibsvmException("The LIBSVM learner cannot predict the next class, because the learning model cannot be found. ");
} else if (featureVector == null) {
throw new LibsvmException("The LIBSVM learner cannot predict the next class, because the feature vector cannot be found. ");
}
int j = 0;
int offset = 0;
int i = 0;
for (FeatureFunction feature : featureVector) {
final FeatureValue featureValue = feature.getFeatureValue();
if (!(excludeNullValues == true && featureValue.isNullValue())) {
if (featureValue instanceof SingleFeatureValue) {
if (((SingleFeatureValue)featureValue).getCode() < cardinalities[i]) {
if (j >= xlist.size()) {
svm_node x = new svm_node();
x.value = 1;
xlist.add(j,x);
}
xlist.get(j++).index = ((SingleFeatureValue)featureValue).getCode() + offset;
}
} else if (featureValue instanceof MultipleFeatureValue) {
for (Integer value : ((MultipleFeatureValue)featureValue).getCodes()) {
if (value < cardinalities[i]) {
// if (((MultipleFeatureValue)featureValue).isKnown(value)) {
if (j >= xlist.size()) {
svm_node x = new svm_node();
x.value = 1;
xlist.add(j,x);
}
xlist.get(j++).index = value + offset;
}
}
}
}
offset += cardinalities[i];
i++;
}
svm_node[] xarray = new svm_node[j];
for (int k = 0; k < j; k++) {
xarray[k] = xlist.get(k);
}
try {
if (decision.getKBestList().getK() == 1 || svm.svm_get_svm_type(model) == svm_parameter.ONE_CLASS ||
svm.svm_get_svm_type(model) == svm_parameter.EPSILON_SVR ||
svm.svm_get_svm_type(model) == svm_parameter.NU_SVR) {
decision.getKBestList().add((int)svm.svm_predict(model, xarray));
} else {
svm_predict_with_kbestlist(model, xarray, decision.getKBestList());
}
} catch (OutOfMemoryError e) {
throw new LibsvmException("Out of memory. Please increase the Java heap size (-Xmx<size>). ", e);
}
return true;
}
public void terminate() throws MaltChainedException {
closeInstanceWriter();
model = null;
svmParam = null;
xlist = null;
owner = null;
}
public BufferedWriter getInstanceWriter() {
return instanceOutput;
}
protected void closeInstanceWriter() throws MaltChainedException {
try {
if (instanceOutput != null) {
instanceOutput.flush();
instanceOutput.close();
instanceOutput = null;
}
} catch (IOException e) {
throw new LibsvmException("The LIBSVM learner cannot close the instance file. ", e);
}
}
/**
* Initialize the LIBSVM according to the parameter string
*
* @param paramString the parameter string to configure the LIBSVM learner.
* @throws MaltChainedException
*/
protected void initSvmParam(String paramString) throws MaltChainedException {
this.paramString = paramString;
svmParam = new svm_parameter();
initParameters(svmParam);
parseParameters(paramString, svmParam);
}
/**
* Returns the parameter string for used for configure LIBSVM
*
* @return the parameter string for used for configure LIBSVM
*/
public String getParamString() {
return paramString;
}
public InstanceModel getOwner() {
return owner;
}
protected void setOwner(InstanceModel owner) {
this.owner = owner;
}
public int getLearnerMode() {
return learnerMode;
}
public void setLearnerMode(int learnerMode) throws MaltChainedException {
this.learnerMode = learnerMode;
}
public String getLearningMethodName() {
return name;
}
/**
* Returns the current configuration
*
* @return the current configuration
* @throws MaltChainedException
*/
public DependencyParserConfig getConfiguration() throws MaltChainedException {
return owner.getGuide().getConfiguration();
}
public int getNumberOfInstances() throws MaltChainedException {
if(numberOfInstances!=0)
return numberOfInstances;
else{
//Do a line count of the instance file and return that
BufferedReader reader = new BufferedReader( getInstanceInputStreamReader(".ins"));
try {
while(reader.readLine()!=null){
numberOfInstances++;
owner.increaseFrequency();
}
reader.close();
} catch (IOException e) {
throw new MaltChainedException("No instances found in file",e);
}
return numberOfInstances;
}
}
public void increaseNumberOfInstances() {
numberOfInstances++;
owner.increaseFrequency();
}
public void decreaseNumberOfInstances() {
numberOfInstances--;
owner.decreaseFrequency();
}
protected void setNumberOfInstances(int numberOfInstances) {
this.numberOfInstances = 0;
}
protected void setLearningMethodName(String name) {
this.name = name;
}
protected OutputStreamWriter getInstanceOutputStreamWriter(String suffix) throws MaltChainedException {
return getConfiguration().getConfigurationDir().getAppendOutputStreamWriter(owner.getModelName()+getLearningMethodName()+suffix);
}
protected InputStreamReader getInstanceInputStreamReader(String suffix) throws MaltChainedException {
return getConfiguration().getConfigurationDir().getInputStreamReader(owner.getModelName()+getLearningMethodName()+suffix);
}
protected InputStreamReader getInstanceInputStreamReaderFromConfigFile(String suffix) throws MaltChainedException {
return getConfiguration().getConfigurationDir().getInputStreamReaderFromConfigFile(owner.getModelName()+getLearningMethodName()+suffix);
}
protected File getFile(String suffix) throws MaltChainedException {
return getConfiguration().getConfigurationDir().getFile(owner.getModelName()+getLearningMethodName()+suffix);
}
protected JarEntry getConfigFileEntry(String suffix) throws MaltChainedException {
return getConfiguration().getConfigurationDir().getConfigFileEntry(owner.getModelName()+getLearningMethodName()+suffix);
}
/**
* Reads an instance file into a svm_problem object according to the Malt-SVM format, which is column fixed format (tab-separated).
*
* @param isr the instance stream reader for the instance file
* @param cardinalities a array containing the number of distinct values for a particular column.
* @param param a svm_parameter object
* @throws LibsvmException
*/
public final svm_problem readProblemMaltSVMFormat(InputStreamReader isr, int[] cardinalities, svm_parameter param) throws MaltChainedException {
final svm_problem prob = new svm_problem();
try {
final BufferedReader fp = new BufferedReader(isr);
int max_index = 0;
if (xlist == null) {
xlist = new ArrayList<svm_node>();
}
prob.l = getNumberOfInstances();
prob.x = new svm_node[prob.l][];
prob.y = new double[prob.l];
int i = 0;
final Pattern tabPattern = Pattern.compile("\t");
final Pattern pipePattern = Pattern.compile("\\|");
while(true) {
String line = fp.readLine();
if(line == null) break;
String[] columns = tabPattern.split(line);
if (columns.length == 0) {
continue;
}
int offset = 0;
int j = 0;
try {
prob.y[i] = (double)Integer.parseInt(columns[j]);
int p = 0;
for(j = 1; j < columns.length; j++) {
final String[] items = pipePattern.split(columns[j]);
for (int k = 0; k < items.length; k++) {
try {
if (Integer.parseInt(items[k]) != -1) {
xlist.add(p, new svm_node());
xlist.get(p).value = 1;
xlist.get(p).index = Integer.parseInt(items[k])+offset;
p++;
}
} catch (NumberFormatException e) {
throw new LibsvmException("The instance file contain a non-integer value '"+items[k]+"'", e);
}
}
offset += cardinalities[j-1];
}
prob.x[i] = xlist.subList(0, p).toArray(new svm_node[0]);
if(columns.length > 1) {
max_index = Math.max(max_index, xlist.get(p-1).index);
}
i++;
xlist.clear();
} catch (ArrayIndexOutOfBoundsException e) {
throw new LibsvmException("Cannot read from the instance file. ", e);
}
}
fp.close();
if (param.gamma == 0) {
param.gamma = 1.0/max_index;
}
xlist = null;
} catch (IOException e) {
throw new LibsvmException("Cannot read from the instance file. ", e);
}
return prob;
}
protected void initSpecialParameters() throws MaltChainedException {
if (getConfiguration().getOptionValue("singlemalt", "null_value") != null && getConfiguration().getOptionValue("singlemalt", "null_value").toString().equalsIgnoreCase("none")) {
excludeNullValues = true;
} else {
excludeNullValues = false;
}
saveInstanceFiles = ((Boolean)getConfiguration().getOptionValue("libsvm", "save_instance_files")).booleanValue();
if (!getConfiguration().getOptionValue("libsvm", "libsvm_external").toString().equals("")) {
try {
if (!new File(getConfiguration().getOptionValue("libsvm", "libsvm_external").toString()).exists()) {
throw new LibsvmException("The path to the external LIBSVM trainer 'svm-train' is wrong.");
}
if (new File(getConfiguration().getOptionValue("libsvm", "libsvm_external").toString()).isDirectory()) {
throw new LibsvmException("The option --libsvm-libsvm_external points to a directory, the path should point at the 'svm-train' file or the 'svm-train.exe' file");
}
if (!(getConfiguration().getOptionValue("libsvm", "libsvm_external").toString().endsWith("svm-train") || getConfiguration().getOptionValue("libsvm", "libsvm_external").toString().endsWith("svm-train.exe"))) {
throw new LibsvmException("The option --libsvm-libsvm_external does not specify the path to 'svm-train' file or the 'svm-train.exe' file. ");
}
pathExternalSVMTrain = getConfiguration().getOptionValue("libsvm", "libsvm_external").toString();
} catch (SecurityException e) {
throw new LibsvmException("Access denied to the file specified by the option --libsvm-libsvm_external. ", e);
}
}
if (getConfiguration().getOptionValue("libsvm", "verbosity") != null) {
verbosity = Verbostity.valueOf(getConfiguration().getOptionValue("libsvm", "verbosity").toString().toUpperCase());
}
}
/**
* Assign a default value to all svm parameters
*
* @param param a svm_parameter object
*/
protected void initParameters(svm_parameter param) throws MaltChainedException {
if (param == null) {
throw new LibsvmException("Svm-parameters cannot be found. ");
}
param.svm_type = svm_parameter.C_SVC;
param.kernel_type = svm_parameter.POLY;
param.degree = 2;
param.gamma = 0.2; // 1/k
param.coef0 = 0;
param.nu = 0.5;
param.cache_size = 100;
param.C = 1;
param.eps = 1.0;
param.p = 0.1;
param.shrinking = 1;
param.probability = 0;
param.nr_weight = 0;
param.weight_label = new int[0];
param.weight = new double[0];
}
/**
* Returns a string containing all svm-parameters of interest
*
* @param param a svm_parameter object
* @return a string containing all svm-parameters of interest
*/
public String toStringParameters(svm_parameter param) {
if (param == null) {
throw new IllegalArgumentException("Svm-parameters cannot be found. ");
}
final StringBuffer sb = new StringBuffer();
final String[] svmtypes = {"C_SVC", "NU_SVC","ONE_CLASS","EPSILON_SVR","NU_SVR"};
final String[] kerneltypes = {"LINEAR", "POLY","RBF","SIGMOID","PRECOMPUTED"};
final DecimalFormat dform = new DecimalFormat("#0.0#");
final DecimalFormatSymbols sym = new DecimalFormatSymbols();
sym.setDecimalSeparator('.');
dform.setDecimalFormatSymbols(sym);
sb.append("LIBSVM SETTINGS\n");
sb.append(" SVM type : " + svmtypes[param.svm_type] + " (" + param.svm_type + ")\n");
sb.append(" Kernel : " + kerneltypes[param.kernel_type] + " (" + param.kernel_type + ")\n");
if (param.kernel_type == svm_parameter.POLY) {
sb.append(" Degree : " + param.degree + "\n");
}
if (param.kernel_type == svm_parameter.POLY || param.kernel_type == svm_parameter.RBF || param.kernel_type == svm_parameter.SIGMOID) {
sb.append(" Gamma : " + dform.format(param.gamma) + "\n");
if (param.kernel_type == svm_parameter.POLY || param.kernel_type == svm_parameter.SIGMOID) {
sb.append(" Coef0 : " + dform.format(param.coef0) + "\n");
}
}
if (param.svm_type == svm_parameter.NU_SVC || param.svm_type == svm_parameter.NU_SVR || param.svm_type == svm_parameter.ONE_CLASS) {
sb.append(" Nu : " + dform.format(param.nu) + "\n");
}
sb.append(" Cache Size : " + dform.format(param.cache_size) + " MB\n");
if (param.svm_type == svm_parameter.C_SVC || param.svm_type == svm_parameter.NU_SVR || param.svm_type == svm_parameter.EPSILON_SVR) {
sb.append(" C : " + dform.format(param.C) + "\n");
}
sb.append(" Eps : " + dform.format(param.eps) + "\n");
if (param.svm_type == svm_parameter.EPSILON_SVR) {
sb.append(" P : " + dform.format(param.p) + "\n");
}
sb.append(" Shrinking : " + param.shrinking + "\n");
sb.append(" Probability : " + param.probability + "\n");
if (param.svm_type == svm_parameter.C_SVC) {
sb.append(" #Weight : " + param.nr_weight + "\n");
if (param.nr_weight > 0) {
sb.append(" Weight labels : ");
for (int i = 0; i < param.nr_weight; i++) {
sb.append(param.weight_label[i]);
if (i != param.nr_weight-1) {
sb.append(", ");
}
}
sb.append("\n");
for (int i = 0; i < param.nr_weight; i++) {
sb.append(dform.format(param.weight));
if (i != param.nr_weight-1) {
sb.append(", ");
}
}
sb.append("\n");
}
}
return sb.toString();
}
public String[] getSVMParamStringArray(svm_parameter param) {
final ArrayList<String> params = new ArrayList<String>();
if (param.svm_type != 0) {
params.add("-s"); params.add(new Integer(param.svm_type).toString());
}
if (param.kernel_type != 2) {
params.add("-t"); params.add(new Integer(param.kernel_type).toString());
}
if (param.degree != 3) {
params.add("-d"); params.add(new Integer(param.degree).toString());
}
params.add("-g"); params.add(new Double(param.gamma).toString());
if (param.coef0 != 0) {
params.add("-r"); params.add(new Double(param.coef0).toString());
}
if (param.nu != 0.5) {
params.add("-n"); params.add(new Double(param.nu).toString());
}
if (param.cache_size != 100) {
params.add("-m"); params.add(new Double(param.cache_size).toString());
}
if (param.C != 1) {
params.add("-c"); params.add(new Double(param.C).toString());
}
if (param.eps != 0.001) {
params.add("-e"); params.add(new Double(param.eps).toString());
}
if (param.p != 0.1) {
params.add("-p"); params.add(new Double(param.p).toString());
}
if (param.shrinking != 1) {
params.add("-h"); params.add(new Integer(param.shrinking).toString());
}
if (param.probability != 0) {
params.add("-b"); params.add(new Integer(param.probability).toString());
}
return params.toArray(new String[params.size()]);
}
/**
* Parses the parameter string. The parameter string must contain parameter and value pairs, which are separated by a blank
* or a underscore. The parameter begins with a character '-' followed by a one-character flag and the value must comply with
* the parameters data type. Some examples:
*
* -s 0 -t 1 -d 2 -g 0.4 -e 0.1
* -s_0_-t_1_-d_2_-g_0.4_-e_0.1
*
* @param paramstring the parameter string
* @param param a svm_parameter object
* @throws LibsvmException
*/
public void parseParameters(String paramstring, svm_parameter param) throws MaltChainedException {
if (param == null) {
throw new LibsvmException("Svm-parameters cannot be found. ");
}
if (paramstring == null) {
return;
}
final String[] argv;
try {
argv = paramstring.split("[_\\p{Blank}]");
} catch (PatternSyntaxException e) {
throw new LibsvmException("Could not split the svm-parameter string '"+paramstring+"'. ", e);
}
for (int i=0; i < argv.length-1; i++) {
if(argv[i].charAt(0) != '-') {
throw new LibsvmException("The argument flag should start with the following character '-', not with "+argv[i].charAt(0));
}
if(++i>=argv.length) {
throw new LibsvmException("The last argument does not have any value. ");
}
try {
switch(argv[i-1].charAt(1)) {
case 's':
param.svm_type = Integer.parseInt(argv[i]);
break;
case 't':
param.kernel_type = Integer.parseInt(argv[i]);
break;
case 'd':
param.degree = Integer.parseInt(argv[i]);
break;
case 'g':
param.gamma = Double.valueOf(argv[i]).doubleValue();
break;
case 'r':
param.coef0 = Double.valueOf(argv[i]).doubleValue();
break;
case 'n':
param.nu = Double.valueOf(argv[i]).doubleValue();
break;
case 'm':
param.cache_size = Double.valueOf(argv[i]).doubleValue();
break;
case 'c':
param.C = Double.valueOf(argv[i]).doubleValue();
break;
case 'e':
param.eps = Double.valueOf(argv[i]).doubleValue();
break;
case 'p':
param.p = Double.valueOf(argv[i]).doubleValue();
break;
case 'h':
param.shrinking = Integer.parseInt(argv[i]);
break;
case 'b':
param.probability = Integer.parseInt(argv[i]);
break;
case 'w':
++param.nr_weight;
{
int[] old = param.weight_label;
param.weight_label = new int[param.nr_weight];
System.arraycopy(old,0,param.weight_label,0,param.nr_weight-1);
}
{
double[] old = param.weight;
param.weight = new double[param.nr_weight];
System.arraycopy(old,0,param.weight,0,param.nr_weight-1);
}
param.weight_label[param.nr_weight-1] = Integer.parseInt(argv[i].substring(2));
param.weight[param.nr_weight-1] = Double.valueOf(argv[i]).doubleValue();
break;
case 'Y':
case 'V':
case 'S':
case 'F':
case 'T':
case 'M':
case 'N':
break;
default:
throw new LibsvmException("Unknown svm parameter: '"+argv[i-1]+"' with value '"+argv[i]+"'. ");
}
} catch (ArrayIndexOutOfBoundsException e) {
throw new LibsvmException("The svm-parameter '"+argv[i-1]+"' could not convert the string value '"+argv[i]+"' into a correct numeric value. ", e);
} catch (NumberFormatException e) {
throw new LibsvmException("The svm-parameter '"+argv[i-1]+"' could not convert the string value '"+argv[i]+"' into a correct numeric value. ", e);
} catch (NullPointerException e) {
throw new LibsvmException("The svm-parameter '"+argv[i-1]+"' could not convert the string value '"+argv[i]+"' into a correct numeric value. ", e);
}
}
}
public void svm_predict_with_kbestlist(svm_model model, svm_node[] x, KBestList kBestList) throws MaltChainedException {
int i;
final int nr_class = svm.svm_get_nr_class(model);
final double[] dec_values = new double[nr_class*(nr_class-1)/2];
svm.svm_predict_values(model, x, dec_values);
final int[] vote = new int[nr_class];
final double[] score = new double[nr_class];
final int[] voteindex = new int[nr_class];
for(i=0;i<nr_class;i++) {
vote[i] = 0;
score[i] = 0.0;
voteindex[i] = i;
}
int pos=0;
for(i=0;i<nr_class;i++) {
for(int j=i+1;j<nr_class;j++) {
if(dec_values[pos] > 0) {
vote[i]++;
} else {
vote[j]++;
}
score[i] += dec_values[pos];
score[j] += dec_values[pos];
pos++;
}
}
for(i=0;i<nr_class;i++) {
score[i] = score[i]/nr_class;
}
int lagest, tmpint;
double tmpdouble;
for (i=0;i<nr_class-1;i++) {
lagest = i;
for (int j=i;j<nr_class;j++) {
if (vote[j] > vote[lagest]) {
lagest = j;
}
}
tmpint = vote[lagest];
vote[lagest] = vote[i];
vote[i] = tmpint;
tmpdouble = score[lagest];
score[lagest] = score[i];
score[i] = tmpdouble;
tmpint = voteindex[lagest];
voteindex[lagest] = voteindex[i];
voteindex[i] = tmpint;
}
final int[] labels = new int[nr_class];
svm.svm_get_labels(model, labels);
int k = nr_class-1;
if (kBestList.getK() != -1) {
k = kBestList.getK() - 1;
}
for (i=0; i<nr_class && k >= 0; i++, k--) {
if (vote[i] > 0 || i == 0) {
if (kBestList instanceof ScoredKBestList) {
((ScoredKBestList)kBestList).add(labels[voteindex[i]], (float)vote[i]/(float)(nr_class*(nr_class-1)/2));
} else {
kBestList.add(labels[voteindex[i]]);
}
}
}
}
/**
* Converts the instance file (Malt's own SVM format) into the LIBSVM (SVMLight) format. The input instance file is removed (replaced)
* by the instance file in the LIBSVM (SVMLight) format. If a column contains -1, the value will be removed in destination file.
*
* @param isr the input stream reader for the source instance file
* @param osw the output stream writer for the destination instance file
* @param cardinalities a vector containing the number of distinct values for a particular column
* @throws LibsvmException
*/
public static void maltSVMFormat2OriginalSVMFormat(InputStreamReader isr, OutputStreamWriter osw, int[] cardinalities) throws MaltChainedException {
try {
final BufferedReader in = new BufferedReader(isr);
final BufferedWriter out = new BufferedWriter(osw);
int c;
int j = 0;
int offset = 0;
int code = 0;
while(true) {
c = in.read();
if (c == -1) {
break;
}
if (c == '\t' || c == '|') {
if (j == 0) {
out.write(Integer.toString(code));
j++;
} else {
if (code != -1) {
out.write(' ');
out.write(Integer.toString(code+offset));
out.write(":1");
}
if (c == '\t') {
offset += cardinalities[j-1];
j++;
}
}
code = 0;
} else if (c == '\n') {
j = 0;
offset = 0;
out.write('\n');
code = 0;
} else if (c == '-') {
code = -1;
} else if (code != -1) {
if (c > 47 && c < 58) {
code = code * 10 + (c-48);
} else {
throw new LibsvmException("The instance file contain a non-integer value, when converting the Malt SVM format into LIBSVM format.");
}
}
}
in.close();
out.close();
} catch (IOException e) {
throw new LibsvmException("Cannot read from the instance file, when converting the Malt SVM format into LIBSVM format. ", e);
}
}
protected void finalize() throws Throwable {
try {
closeInstanceWriter();
} finally {
super.finalize();
}
}
/* (non-Javadoc)
* @see java.lang.Object#toString()
*/
public String toString() {
final StringBuffer sb = new StringBuffer();
sb.append("\nLIBSVM INTERFACE\n");
sb.append(" LIBSVM version: "+LIBSVM_VERSION+"\n");
sb.append(" SVM-param string: "+paramString+"\n");
sb.append(toStringParameters(svmParam));
return sb.toString();
}
@Override
public void divideByFeatureSet(
Set<Integer> featureIdsToCreateSeparateBranchesForSet, ArrayList<Integer> divideFeatureIndexVector, String otherId) throws MaltChainedException {
//Create a hash map that maps every feature id to a writer
HashMap<Integer, BufferedWriter> featureIdToWriterMap = new HashMap<Integer, BufferedWriter>();
for(int element:featureIdsToCreateSeparateBranchesForSet){
BufferedWriter outputWriter = new BufferedWriter(getConfiguration().getConfigurationDir().getOutputStreamWriter(owner.getModelName().replace('.','_') + element + "." + getLearningMethodName()+".ins"));
featureIdToWriterMap.put(element, outputWriter);
}
BufferedWriter otherOutputWriter = new BufferedWriter(getConfiguration().getConfigurationDir().getOutputStreamWriter(owner.getModelName().replace('.','_') + otherId + "." + getLearningMethodName()+".ins"));
try {
final BufferedReader in = new BufferedReader(getInstanceInputStreamReader(".ins"));
//every line will be written to a separate file
String line = in.readLine();
final Pattern tabPattern = Pattern.compile("\t");
while(line!=null){
//Find out which pot the line shall be put in
String[] lineArray = tabPattern.split(line);
int id = new Integer(lineArray[divideFeatureIndexVector.get(0)+1]);
if(!featureIdToWriterMap.containsKey(id)){
otherOutputWriter.write(line + "\n");
}else
featureIdToWriterMap.get(id).write(getLineToWrite(lineArray,divideFeatureIndexVector.get(0)+1));
line = in.readLine();
}
otherOutputWriter.close();
in.close();
for(BufferedWriter writer: featureIdToWriterMap.values())
writer.close();
} catch (SecurityException e) {
throw new LiblinearException("The Liblinear learner cannot remove the instance file. ", e);
} catch (NullPointerException e) {
throw new LiblinearException("The instance file cannot be found. ", e);
} catch (FileNotFoundException e) {
throw new LiblinearException("The instance file cannot be found. ", e);
} catch (IOException e) {
throw new LiblinearException("The Liblinear learner read from the instance file. ", e);
}
}
private String getLineToWrite(String[] lineArray, int excludeIndex) {
StringBuffer buf = new StringBuffer();
for(int n = 0; n < lineArray.length; n++)
if(n != excludeIndex)
buf.append(lineArray[n] + "\t");
buf.append("\n");
return buf.toString();
}
@Override
public Map<Integer, Integer> createFeatureIdToCountMap(
ArrayList<Integer> divideFeatureIndexVector) throws MaltChainedException{
HashMap<Integer, Integer> featureIdToCountMap = new HashMap<Integer, Integer>();
//Go trough the file and count all feature ids in the given column(s)
try {
final BufferedReader in = new BufferedReader(getInstanceInputStreamReader(".ins"));
//every line will be written to a separate file
String line = in.readLine();
final Pattern tabPattern = Pattern.compile("\t");
while(line!=null){
//Find out which pot the line shall be put in
String[] lineArray = tabPattern.split(line);
for(int n = 0; n < divideFeatureIndexVector.size(); n++){
int id = new Integer(lineArray[divideFeatureIndexVector.get(n)+1]);
if (!featureIdToCountMap.containsKey(id)) {
featureIdToCountMap.put(id, 0);
}
int previousCount = featureIdToCountMap.get(id);
featureIdToCountMap.put(id, previousCount + 1);
}
line = in.readLine();
}
in.close();
} catch (SecurityException e) {
throw new LiblinearException("The Libsvm learner cannot remove the instance file. ", e);
} catch (NullPointerException e) {
throw new LiblinearException("The instance file cannot be found. ", e);
} catch (FileNotFoundException e) {
throw new LiblinearException("The instance file cannot be found. ", e);
} catch (IOException e) {
throw new LiblinearException("The Liblinear learner read from the instance file. ", e);
}
return featureIdToCountMap;
}
}