/**
* Run the job
*
* @param params The Job parameters containing the gramSize, input output folders, defaultCat,
* encoding
*/
public static void runJob(Parameters params) throws IOException {
Configurable client = new JobClient();
JobConf conf = new JobConf(BayesClassifierDriver.class);
conf.setJobName("Bayes Classifier Driver running over input: " + params.get("testDirPath"));
conf.setOutputKeyClass(StringTuple.class);
conf.setOutputValueClass(DoubleWritable.class);
FileInputFormat.setInputPaths(conf, new Path(params.get("testDirPath")));
Path outPath = new Path(params.get("testDirPath") + "-output");
FileOutputFormat.setOutputPath(conf, outPath);
conf.setInputFormat(KeyValueTextInputFormat.class);
conf.setMapperClass(BayesClassifierMapper.class);
conf.setCombinerClass(BayesClassifierReducer.class);
conf.setReducerClass(BayesClassifierReducer.class);
conf.setOutputFormat(SequenceFileOutputFormat.class);
conf.set(
"io.serializations",
"org.apache.hadoop.io.serializer.JavaSerialization,"
+ "org.apache.hadoop.io.serializer.WritableSerialization");
HadoopUtil.overwriteOutput(outPath);
conf.set("bayes.parameters", params.toString());
client.setConf(conf);
JobClient.runJob(conf);
Path outputFiles = new Path(outPath, "part*");
FileSystem dfs = FileSystem.get(outPath.toUri(), conf);
ConfusionMatrix matrix = readResult(dfs, outputFiles, conf, params);
log.info("{}", matrix.summarize());
}
/**
* @param args
* @throws Exception
*/
public static void main(String args[]) throws Exception {
log.setLevel(Level.INFO);
Configuration v_HadoopConf = new Configuration();
s_HadoopConf = v_HadoopConf;
s_Config = Config.getInstance();
String v_PathPrefix = s_Config.getBaseDir() + s_Config.getDataDir() + "mahout/";
String v_TextDir = s_Config.getTextPath();
Path m_DocumentDir = new Path(v_TextDir);
Path m_SequenceDir = new Path(v_PathPrefix, "sequence/");
Path m_TokensDir = new Path(v_PathPrefix, "tokens");
Path m_TF = new Path(v_PathPrefix, "termfreq/");
String m_VecFolder = "Vectors";
Path m_tf_idf = new Path(v_PathPrefix, "tfidf/");
boolean m_Sequential = true;
HadoopUtil.delete(v_HadoopConf, new Path(v_PathPrefix, "clusters/"));
if (!s_Config.getReuseTFIDF()) {
createTFIDF(
v_HadoopConf, m_DocumentDir, m_SequenceDir, m_TokensDir, m_TF, m_VecFolder, m_tf_idf);
}
HierarchicalKMeansClusterer v_Hkmc = new HierarchicalKMeansClusterer();
SetTree<ClusterDescriptor> v_Tree = v_Hkmc.run(s_HadoopConf, m_Sequential);
saveAsTree(v_Tree);
saveAsXml(v_Tree);
}
@Override
public int run(String[] args) throws Exception {
addInputOption();
addOutputOption();
addOption(DefaultOptionCreator.distanceMeasureOption().create());
addOption(DefaultOptionCreator.t1Option().create());
addOption(DefaultOptionCreator.t2Option().create());
addOption(DefaultOptionCreator.overwriteOption().create());
Map<String, List<String>> argMap = parseArguments(args);
if (argMap == null) {
return -1;
}
Path input = getInputPath();
Path output = getOutputPath();
if (hasOption(DefaultOptionCreator.OVERWRITE_OPTION)) {
HadoopUtil.delete(new Configuration(), output);
}
String measureClass = getOption(DefaultOptionCreator.DISTANCE_MEASURE_OPTION);
double t1 = Double.parseDouble(getOption(DefaultOptionCreator.T1_OPTION));
double t2 = Double.parseDouble(getOption(DefaultOptionCreator.T2_OPTION));
DistanceMeasure measure = ClassUtils.instantiateAs(measureClass, DistanceMeasure.class);
run(input, output, measure, t1, t2);
return 0;
}
/**
* Count the document frequencies of features in parallel using Map/Reduce. The input documents
* have to be in {@link SequenceFile} format
*/
private static void startDFCounting(Path input, Path output, Configuration baseConf)
throws IOException, InterruptedException, ClassNotFoundException {
Configuration conf = new Configuration(baseConf);
// this conf parameter needs to be set enable serialisation of conf values
conf.set(
"io.serializations",
"org.apache.hadoop.io.serializer.JavaSerialization,"
+ "org.apache.hadoop.io.serializer.WritableSerialization");
Job job = new Job(conf);
job.setJobName("VectorTfIdf Document Frequency Count running over input: " + input);
job.setJarByClass(TFIDFConverter.class);
job.setOutputKeyClass(IntWritable.class);
job.setOutputValueClass(LongWritable.class);
FileInputFormat.setInputPaths(job, input);
FileOutputFormat.setOutputPath(job, output);
job.setMapperClass(TermDocumentCountMapper.class);
job.setInputFormatClass(SequenceFileInputFormat.class);
job.setCombinerClass(TermDocumentCountReducer.class);
job.setReducerClass(TermDocumentCountReducer.class);
job.setOutputFormatClass(SequenceFileOutputFormat.class);
HadoopUtil.delete(conf, output);
boolean succeeded = job.waitForCompletion(true);
if (!succeeded) {
throw new IllegalStateException("Job failed!");
}
}
@Override
protected void setup(Context ctx) throws IOException, InterruptedException {
try {
Configuration conf = ctx.getConfiguration();
Path[] localFiles = DistributedCache.getLocalCacheFiles(conf);
Preconditions.checkArgument(
localFiles != null && localFiles.length >= 1,
"missing paths from the DistributedCache");
Path inputVectorPath = HadoopUtil.getSingleCachedFile(conf);
SequenceFileValueIterator<VectorWritable> iterator =
new SequenceFileValueIterator<VectorWritable>(inputVectorPath, true, conf);
try {
inputVector = iterator.next().get();
} finally {
Closeables.close(iterator, true);
}
int outDim = conf.getInt(OUTPUT_VECTOR_DIMENSION, Integer.MAX_VALUE);
outputVector =
conf.getBoolean(IS_SPARSE_OUTPUT, false)
? new RandomAccessSparseVector(outDim, 10)
: new DenseVector(outDim);
} catch (IOException ioe) {
throw new IllegalStateException(ioe);
}
}
/**
* Create Term Frequency-Inverse Document Frequency (Tf-Idf) Vectors from the input set of vectors
* in {@link SequenceFile} format. This job uses a fixed limit on the maximum memory used by the
* feature chunk per node thereby splitting the process across multiple map/reduces. Before using
* this method calculateDF should be called
*
* @param input input directory of the vectors in {@link SequenceFile} format
* @param output output directory where {@link org.apache.mahout.math.RandomAccessSparseVector}'s
* of the document are generated
* @param datasetFeatures Document frequencies information calculated by calculateDF
* @param minDf The minimum document frequency. Default 1
* @param maxDF The max percentage of vectors for the DF. Can be used to remove really high
* frequency features. Expressed as an integer between 0 and 100. Default 99
* @param numReducers The number of reducers to spawn. This also affects the possible parallelism
* since each reducer will typically produce a single output file containing tf-idf vectors
* for a subset of the documents in the corpus.
*/
public static void processTfIdf(
Path input,
Path output,
Configuration baseConf,
Pair<Long[], List<Path>> datasetFeatures,
int minDf,
long maxDF,
float normPower,
boolean logNormalize,
boolean sequentialAccessOutput,
boolean namedVector,
int numReducers)
throws IOException, InterruptedException, ClassNotFoundException {
Preconditions.checkArgument(
normPower == PartialVectorMerger.NO_NORMALIZING || normPower >= 0,
"If specified normPower must be nonnegative",
normPower);
Preconditions.checkArgument(
normPower == PartialVectorMerger.NO_NORMALIZING
|| (normPower > 1 && !Double.isInfinite(normPower))
|| !logNormalize,
"normPower must be > 1 and not infinite if log normalization is chosen",
normPower);
int partialVectorIndex = 0;
List<Path> partialVectorPaths = Lists.newArrayList();
List<Path> dictionaryChunks = datasetFeatures.getSecond();
for (Path dictionaryChunk : dictionaryChunks) {
Path partialVectorOutputPath = new Path(output, VECTOR_OUTPUT_FOLDER + partialVectorIndex++);
partialVectorPaths.add(partialVectorOutputPath);
makePartialVectors(
input,
baseConf,
datasetFeatures.getFirst()[0],
datasetFeatures.getFirst()[1],
minDf,
maxDF,
dictionaryChunk,
partialVectorOutputPath,
sequentialAccessOutput,
namedVector);
}
Configuration conf = new Configuration(baseConf);
Path outputDir = new Path(output, DOCUMENT_VECTOR_OUTPUT_FOLDER);
PartialVectorMerger.mergePartialVectors(
partialVectorPaths,
outputDir,
baseConf,
normPower,
logNormalize,
datasetFeatures.getFirst()[0].intValue(),
sequentialAccessOutput,
namedVector,
numReducers);
HadoopUtil.delete(conf, partialVectorPaths);
}
/**
* Merge all the partial {@link org.apache.mahout.math.RandomAccessSparseVector}s into the
* complete Document {@link org.apache.mahout.math.RandomAccessSparseVector}
*
* @param partialVectorPaths input directory of the vectors in {@link
* org.apache.hadoop.io.SequenceFile} format
* @param output output directory were the partial vectors have to be created
* @param baseConf job configuration
* @param normPower The normalization value. Must be greater than or equal to 0 or equal to {@link
* #NO_NORMALIZING}
* @param dimension
* @param sequentialAccess output vectors should be optimized for sequential access
* @param namedVector output vectors should be named, retaining key (doc id) as a label
* @param numReducers The number of reducers to spawn
*/
public static void mergePartialVectors(
Iterable<Path> partialVectorPaths,
Path output,
Configuration baseConf,
float normPower,
boolean logNormalize,
int dimension,
boolean sequentialAccess,
boolean namedVector,
int numReducers)
throws IOException, InterruptedException, ClassNotFoundException {
Preconditions.checkArgument(
normPower == NO_NORMALIZING || normPower >= 0,
"If specified normPower must be nonnegative",
normPower);
Preconditions.checkArgument(
normPower == NO_NORMALIZING
|| (normPower > 1 && !Double.isInfinite(normPower))
|| !logNormalize,
"normPower must be > 1 and not infinite if log normalization is chosen",
normPower);
Configuration conf = new Configuration(baseConf);
// this conf parameter needs to be set enable serialisation of conf values
conf.set(
"io.serializations",
"org.apache.hadoop.io.serializer.JavaSerialization,"
+ "org.apache.hadoop.io.serializer.WritableSerialization");
conf.setBoolean(SEQUENTIAL_ACCESS, sequentialAccess);
conf.setBoolean(NAMED_VECTOR, namedVector);
conf.setInt(DIMENSION, dimension);
conf.setFloat(NORMALIZATION_POWER, normPower);
conf.setBoolean(LOG_NORMALIZE, logNormalize);
Job job = new Job(conf);
job.setJobName("PartialVectorMerger::MergePartialVectors");
job.setJarByClass(PartialVectorMerger.class);
job.setOutputKeyClass(Text.class);
job.setOutputValueClass(VectorWritable.class);
FileInputFormat.setInputPaths(job, getCommaSeparatedPaths(partialVectorPaths));
FileOutputFormat.setOutputPath(job, output);
job.setMapperClass(Mapper.class);
job.setInputFormatClass(SequenceFileInputFormat.class);
job.setReducerClass(PartialVectorMergeReducer.class);
job.setOutputFormatClass(SequenceFileOutputFormat.class);
job.setNumReduceTasks(numReducers);
HadoopUtil.delete(conf, output);
boolean succeeded = job.waitForCompletion(true);
if (!succeeded) {
throw new IllegalStateException("Job failed!");
}
}
public static void main(String[] args) throws Exception {
if (args.length > 0) {
log.info("Running with only user-supplied arguments");
ToolRunner.run(new Configuration(), new Job(), args);
} else {
log.info("Running with default arguments");
Path output = new Path("output");
HadoopUtil.delete(new Configuration(), output);
run(new Path("testdata"), output, new EuclideanDistanceMeasure(), 80, 55);
}
}
public static Job createTimesSquaredJob(
Configuration initialConf,
Vector v,
int outputVectorDim,
Path matrixInputPath,
Path outputVectorPathBase,
Class<? extends TimesSquaredMapper> mapClass,
Class<? extends VectorSummingReducer> redClass)
throws IOException {
FileSystem fs = FileSystem.get(matrixInputPath.toUri(), initialConf);
matrixInputPath = fs.makeQualified(matrixInputPath);
outputVectorPathBase = fs.makeQualified(outputVectorPathBase);
long now = System.nanoTime();
Path inputVectorPath = new Path(outputVectorPathBase, INPUT_VECTOR + '/' + now);
SequenceFile.Writer inputVectorPathWriter = null;
try {
inputVectorPathWriter =
new SequenceFile.Writer(
fs, initialConf, inputVectorPath, NullWritable.class, VectorWritable.class);
inputVectorPathWriter.append(NullWritable.get(), new VectorWritable(v));
} finally {
Closeables.close(inputVectorPathWriter, false);
}
URI ivpURI = inputVectorPath.toUri();
DistributedCache.setCacheFiles(new URI[] {ivpURI}, initialConf);
Job job =
HadoopUtil.prepareJob(
matrixInputPath,
new Path(outputVectorPathBase, OUTPUT_VECTOR_FILENAME),
SequenceFileInputFormat.class,
mapClass,
NullWritable.class,
VectorWritable.class,
redClass,
NullWritable.class,
VectorWritable.class,
SequenceFileOutputFormat.class,
initialConf);
job.setCombinerClass(redClass);
job.setJobName("TimesSquaredJob: " + matrixInputPath);
Configuration conf = job.getConfiguration();
conf.set(INPUT_VECTOR, ivpURI.toString());
conf.setBoolean(IS_SPARSE_OUTPUT, !v.isDense());
conf.setInt(OUTPUT_VECTOR_DIMENSION, outputVectorDim);
return job;
}
/**
* Run the job
*
* @param input the input pathname String
* @param output the output pathname String
* @param catFile the file containing the Wikipedia categories
* @param exactMatchOnly if true, then the Wikipedia category must match exactly instead of simply
* containing the category string
*/
public static void runJob(
String input,
String output,
String catFile,
boolean exactMatchOnly,
Class<? extends Analyzer> analyzerClass)
throws IOException, InterruptedException, ClassNotFoundException {
Configuration conf = new Configuration();
conf.set("key.value.separator.in.input.line", " ");
conf.set("xmlinput.start", "<page>");
conf.set("xmlinput.end", "</page>");
conf.setBoolean("exact.match.only", exactMatchOnly);
conf.set("analyzer.class", analyzerClass.getName());
conf.set(
"io.serializations",
"org.apache.hadoop.io.serializer.JavaSerialization,"
+ "org.apache.hadoop.io.serializer.WritableSerialization");
// Dont ever forget this. People should keep track of how hadoop conf
// parameters can make or break a piece of code
Set<String> categories = Sets.newHashSet();
for (String line : new FileLineIterable(new File(catFile))) {
categories.add(line.trim().toLowerCase(Locale.ENGLISH));
}
Stringifier<Set<String>> setStringifier =
new DefaultStringifier<Set<String>>(conf, GenericsUtil.getClass(categories));
String categoriesStr = setStringifier.toString(categories);
conf.set("wikipedia.categories", categoriesStr);
Job job = new Job(conf);
log.info("Input: {} Out: {} Categories: {}", input, output, catFile);
job.setJarByClass(WikipediaDatasetCreatorDriver.class);
job.setOutputKeyClass(Text.class);
job.setOutputValueClass(Text.class);
job.setMapperClass(WikipediaDatasetCreatorMapper.class);
// TODO: job.setNumMapTasks(100);
job.setInputFormatClass(XmlInputFormat.class);
job.setReducerClass(WikipediaDatasetCreatorReducer.class);
job.setOutputFormatClass(TextOutputFormat.class);
FileInputFormat.setInputPaths(job, new Path(input));
Path outPath = new Path(output);
FileOutputFormat.setOutputPath(job, outPath);
HadoopUtil.delete(conf, outPath);
boolean succeeded = job.waitForCompletion(true);
if (!succeeded) {
throw new IllegalStateException("Job failed!");
}
}
/**
* Run the job
*
* @param input the input pathname String
* @param output the output pathname String
* @param catFile the file containing the Wikipedia categories
* @param exactMatchOnly if true, then the Wikipedia category must match exactly instead of simply
* containing the category string
* @param all if true select all categories
* @throws ClassNotFoundException
* @throws InterruptedException
*/
public static void runJob(
String input, String output, String catFile, boolean exactMatchOnly, boolean all)
throws IOException, InterruptedException, ClassNotFoundException {
Configuration conf = new Configuration();
conf.set("xmlinput.start", "<page>");
conf.set("xmlinput.end", "</page>");
conf.setBoolean("exact.match.only", exactMatchOnly);
conf.setBoolean("all.files", all);
conf.set(
"io.serializations",
"org.apache.hadoop.io.serializer.JavaSerialization,"
+ "org.apache.hadoop.io.serializer.WritableSerialization");
Job job = new Job(conf);
if (WikipediaToSequenceFile.log.isInfoEnabled()) {
log.info(
"Input: " + input + " Out: " + output + " Categories: " + catFile + " All Files: " + all);
}
job.setOutputKeyClass(Text.class);
job.setOutputValueClass(Text.class);
FileInputFormat.setInputPaths(job, new Path(input));
Path outPath = new Path(output);
FileOutputFormat.setOutputPath(job, outPath);
job.setMapperClass(WikipediaMapper.class);
job.setInputFormatClass(XmlInputFormat.class);
job.setReducerClass(Reducer.class);
job.setOutputFormatClass(SequenceFileOutputFormat.class);
job.setJarByClass(WikipediaToSequenceFile.class);
/*
* conf.set("mapred.compress.map.output", "true"); conf.set("mapred.map.output.compression.type",
* "BLOCK"); conf.set("mapred.output.compress", "true"); conf.set("mapred.output.compression.type",
* "BLOCK"); conf.set("mapred.output.compression.codec", "org.apache.hadoop.io.compress.GzipCodec");
*/
HadoopUtil.overwriteOutput(outPath);
Set<String> categories = new HashSet<String>();
if (catFile.length() > 0) {
for (String line : new FileLineIterable(new File(catFile))) {
categories.add(line.trim().toLowerCase(Locale.ENGLISH));
}
}
DefaultStringifier<Set<String>> setStringifier =
new DefaultStringifier<Set<String>>(conf, GenericsUtil.getClass(categories));
String categoriesStr = setStringifier.toString(categories);
conf.set("wikipedia.categories", categoriesStr);
job.waitForCompletion(true);
}
/**
* Create a partial tfidf vector using a chunk of features from the input vectors. The input
* vectors has to be in the {@link SequenceFile} format
*
* @param input input directory of the vectors in {@link SequenceFile} format
* @param featureCount Number of unique features in the dataset
* @param vectorCount Number of vectors in the dataset
* @param minDf The minimum document frequency. Default 1
* @param maxDF The max percentage of vectors for the DF. Can be used to remove really high
* frequency features. Expressed as an integer between 0 and 100. Default 99
* @param dictionaryFilePath location of the chunk of features and the id's
* @param output output directory were the partial vectors have to be created
* @param sequentialAccess output vectors should be optimized for sequential access
* @param namedVector output vectors should be named, retaining key (doc id) as a label
*/
private static void makePartialVectors(
Path input,
Configuration baseConf,
Long featureCount,
Long vectorCount,
int minDf,
long maxDF,
Path dictionaryFilePath,
Path output,
boolean sequentialAccess,
boolean namedVector)
throws IOException, InterruptedException, ClassNotFoundException {
Configuration conf = new Configuration(baseConf);
// this conf parameter needs to be set enable serialisation of conf values
conf.set(
"io.serializations",
"org.apache.hadoop.io.serializer.JavaSerialization,"
+ "org.apache.hadoop.io.serializer.WritableSerialization");
conf.setLong(FEATURE_COUNT, featureCount);
conf.setLong(VECTOR_COUNT, vectorCount);
conf.setInt(MIN_DF, minDf);
conf.setLong(MAX_DF, maxDF);
conf.setBoolean(PartialVectorMerger.SEQUENTIAL_ACCESS, sequentialAccess);
conf.setBoolean(PartialVectorMerger.NAMED_VECTOR, namedVector);
DistributedCache.addCacheFile(dictionaryFilePath.toUri(), conf);
Job job = new Job(conf);
job.setJobName(
": MakePartialVectors: input-folder: "
+ input
+ ", dictionary-file: "
+ dictionaryFilePath.toString());
job.setJarByClass(TFIDFConverter.class);
job.setOutputKeyClass(Text.class);
job.setOutputValueClass(VectorWritable.class);
FileInputFormat.setInputPaths(job, input);
FileOutputFormat.setOutputPath(job, output);
job.setMapperClass(Mapper.class);
job.setInputFormatClass(SequenceFileInputFormat.class);
job.setReducerClass(TFIDFPartialVectorReducer.class);
job.setOutputFormatClass(SequenceFileOutputFormat.class);
HadoopUtil.delete(conf, output);
boolean succeeded = job.waitForCompletion(true);
if (!succeeded) {
throw new IllegalStateException("Job failed!");
}
}
private static void createTFIDF(
Configuration p_HadoopConf,
Path p_DocumentDir,
Path p_SequenceDir,
Path p_TokensDir,
Path p_TF,
String p_VecFolder,
Path p_tf_idf)
throws IOException, Exception, ClassNotFoundException, InterruptedException {
log.info(new Date() + " - Creating TF-IDF vectors started.");
HadoopUtil.delete(p_HadoopConf, p_SequenceDir);
HadoopUtil.delete(p_HadoopConf, p_TokensDir);
HadoopUtil.delete(p_HadoopConf, p_TF);
createSequence(p_DocumentDir.toString(), p_SequenceDir.toString());
createTokens(p_SequenceDir, p_TokensDir, s_HadoopConf);
createTFVectors(p_TokensDir, p_TF, p_VecFolder);
createTFIDFVectors(new Path(p_TF, p_VecFolder), p_tf_idf, s_HadoopConf);
log.info("Creating TF-IDF vectors finished.");
}
@Override
public int run(String[] args) throws IOException, ClassNotFoundException, InterruptedException {
addInputOption();
addOutputOption();
addOption(MinhashOptionCreator.minClusterSizeOption().create());
addOption(MinhashOptionCreator.minVectorSizeOption().create());
addOption(MinhashOptionCreator.hashTypeOption().create());
addOption(MinhashOptionCreator.numHashFunctionsOption().create());
addOption(MinhashOptionCreator.keyGroupsOption().create());
addOption(MinhashOptionCreator.numReducersOption().create());
addOption(MinhashOptionCreator.debugOutputOption().create());
addOption(DefaultOptionCreator.overwriteOption().create());
if (parseArguments(args) == null) {
return -1;
}
Path input = getInputPath();
Path output = getOutputPath();
if (hasOption(DefaultOptionCreator.OVERWRITE_OPTION)) {
HadoopUtil.delete(getConf(), output);
}
int minClusterSize = Integer.valueOf(getOption(MinhashOptionCreator.MIN_CLUSTER_SIZE));
int minVectorSize = Integer.valueOf(getOption(MinhashOptionCreator.MIN_VECTOR_SIZE));
String hashType = getOption(MinhashOptionCreator.HASH_TYPE);
int numHashFunctions = Integer.valueOf(getOption(MinhashOptionCreator.NUM_HASH_FUNCTIONS));
int keyGroups = Integer.valueOf(getOption(MinhashOptionCreator.KEY_GROUPS));
int numReduceTasks = Integer.parseInt(getOption(MinhashOptionCreator.NUM_REDUCERS));
boolean debugOutput = hasOption(MinhashOptionCreator.DEBUG_OUTPUT);
runJob(
input,
output,
minClusterSize,
minVectorSize,
hashType,
numHashFunctions,
keyGroups,
numReduceTasks,
debugOutput);
return 0;
}
@Override
public int run(String[] args) throws Exception {
String path = System.getProperty("user.dir");
addInputOption();
addOutputOption();
addOption(ALPHA_I, "a", "smoothing parameter", String.valueOf(1.0f));
addOption(
buildOption(
TRAIN_COMPLEMENTARY, "c", "train complementary?", false, false, String.valueOf(false)));
addOption(LABEL_INDEX, "li", "The path to store the label index in", false);
addOption(DefaultOptionCreator.overwriteOption().create());
Path labPath = new Path(path + "/../out/labelindex/");
long labelSize = createLabelIndex(labPath);
float alphaI = 1.0F;
boolean trainComplementary = true;
HadoopUtil.setSerializations(getConf());
HadoopUtil.cacheFiles(labPath, getConf());
HadoopUtil.delete(getConf(), new Path("/tmp/summedObservations"));
HadoopUtil.delete(getConf(), new Path("/tmp/weights"));
HadoopUtil.delete(getConf(), new Path("/tmp/thetas"));
// Add up all the vectors with the same labels, while mapping the labels into our index
Job indexInstances =
prepareJob(
new Path(path + "/../out/training"),
new Path("/tmp/summedObservations"),
SequenceFileInputFormat.class,
IndexInstancesMapper.class,
IntWritable.class,
VectorWritable.class,
VectorSumReducer.class,
IntWritable.class,
VectorWritable.class,
SequenceFileOutputFormat.class);
indexInstances.setCombinerClass(VectorSumReducer.class);
boolean succeeded = indexInstances.waitForCompletion(true);
if (!succeeded) {
return -1;
}
// Sum up all the weights from the previous step, per label and per feature
Job weightSummer =
prepareJob(
new Path("/tmp/summedObservations"),
new Path("/tmp/weights"),
SequenceFileInputFormat.class,
WeightsMapper.class,
Text.class,
VectorWritable.class,
VectorSumReducer.class,
Text.class,
VectorWritable.class,
SequenceFileOutputFormat.class);
weightSummer.getConfiguration().set(WeightsMapper.NUM_LABELS, String.valueOf(labelSize));
weightSummer.setCombinerClass(VectorSumReducer.class);
succeeded = weightSummer.waitForCompletion(true);
if (!succeeded) {
return -1;
}
// Put the per label and per feature vectors into the cache
HadoopUtil.cacheFiles(new Path("/tmp/weights"), getConf());
if (trainComplementary) {
// Calculate the per label theta normalizers, write out to LABEL_THETA_NORMALIZER vector
// see http://people.csail.mit.edu/jrennie/papers/icml03-nb.pdf - Section 3.2, Weight
// Magnitude Errors
Job thetaSummer =
prepareJob(
new Path("/tmp/summedObservations"),
new Path("/tmp/thetas"),
SequenceFileInputFormat.class,
ThetaMapper.class,
Text.class,
VectorWritable.class,
VectorSumReducer.class,
Text.class,
VectorWritable.class,
SequenceFileOutputFormat.class);
thetaSummer.setCombinerClass(VectorSumReducer.class);
thetaSummer.getConfiguration().setFloat(ThetaMapper.ALPHA_I, alphaI);
thetaSummer
.getConfiguration()
.setBoolean(ThetaMapper.TRAIN_COMPLEMENTARY, trainComplementary);
succeeded = thetaSummer.waitForCompletion(true);
if (!succeeded) {
return -1;
}
}
// Put the per label theta normalizers into the cache
HadoopUtil.cacheFiles(new Path("/tmp/thetas"), getConf());
// Validate our model and then write it out to the official output
getConf().setFloat(ThetaMapper.ALPHA_I, alphaI);
getConf().setBoolean(NaiveBayesModel.COMPLEMENTARY_MODEL, trainComplementary);
NaiveBayesModel naiveBayesModel = BayesUtils.readModelFromDir(new Path("/tmp/"), getConf());
naiveBayesModel.validate();
naiveBayesModel.serialize(new Path(path + "/../out/model"), getConf());
return 0;
}
@Override
public int run(String[] args) throws Exception {
addInputOption();
addOutputOption();
addOption("numberOfColumns", "r", "Number of columns in the input matrix", false);
addOption(
"similarityClassname",
"s",
"Name of distributed similarity class to instantiate, alternatively use "
+ "one of the predefined similarities ("
+ VectorSimilarityMeasures.list()
+ ')');
addOption(
"maxSimilaritiesPerRow",
"m",
"Number of maximum similarities per row (default: "
+ DEFAULT_MAX_SIMILARITIES_PER_ROW
+ ')',
String.valueOf(DEFAULT_MAX_SIMILARITIES_PER_ROW));
addOption(
"excludeSelfSimilarity",
"ess",
"compute similarity of rows to themselves?",
String.valueOf(false));
addOption("threshold", "tr", "discard row pairs with a similarity value below this", false);
addOption(DefaultOptionCreator.overwriteOption().create());
Map<String, List<String>> parsedArgs = parseArguments(args);
if (parsedArgs == null) {
return -1;
}
int numberOfColumns;
if (hasOption("numberOfColumns")) {
// Number of columns explicitly specified via CLI
numberOfColumns = Integer.parseInt(getOption("numberOfColumns"));
} else {
// else get the number of columns by determining the cardinality of a vector in the input
// matrix
numberOfColumns = getDimensions(getInputPath());
}
String similarityClassnameArg = getOption("similarityClassname");
String similarityClassname;
try {
similarityClassname = VectorSimilarityMeasures.valueOf(similarityClassnameArg).getClassname();
} catch (IllegalArgumentException iae) {
similarityClassname = similarityClassnameArg;
}
// Clear the output and temp paths if the overwrite option has been set
if (hasOption(DefaultOptionCreator.OVERWRITE_OPTION)) {
// Clear the temp path
HadoopUtil.delete(getConf(), getTempPath());
// Clear the output path
HadoopUtil.delete(getConf(), getOutputPath());
}
int maxSimilaritiesPerRow = Integer.parseInt(getOption("maxSimilaritiesPerRow"));
boolean excludeSelfSimilarity = Boolean.parseBoolean(getOption("excludeSelfSimilarity"));
double threshold =
hasOption("threshold") ? Double.parseDouble(getOption("threshold")) : NO_THRESHOLD;
Path weightsPath = getTempPath("weights");
Path normsPath = getTempPath("norms.bin");
Path numNonZeroEntriesPath = getTempPath("numNonZeroEntries.bin");
Path maxValuesPath = getTempPath("maxValues.bin");
Path pairwiseSimilarityPath = getTempPath("pairwiseSimilarity");
AtomicInteger currentPhase = new AtomicInteger();
if (shouldRunNextPhase(parsedArgs, currentPhase)) {
Job normsAndTranspose =
prepareJob(
getInputPath(),
weightsPath,
VectorNormMapper.class,
IntWritable.class,
VectorWritable.class,
MergeVectorsReducer.class,
IntWritable.class,
VectorWritable.class);
normsAndTranspose.setCombinerClass(MergeVectorsCombiner.class);
Configuration normsAndTransposeConf = normsAndTranspose.getConfiguration();
normsAndTransposeConf.set(THRESHOLD, String.valueOf(threshold));
normsAndTransposeConf.set(NORMS_PATH, normsPath.toString());
normsAndTransposeConf.set(NUM_NON_ZERO_ENTRIES_PATH, numNonZeroEntriesPath.toString());
normsAndTransposeConf.set(MAXVALUES_PATH, maxValuesPath.toString());
normsAndTransposeConf.set(SIMILARITY_CLASSNAME, similarityClassname);
boolean succeeded = normsAndTranspose.waitForCompletion(true);
if (!succeeded) {
return -1;
}
}
if (shouldRunNextPhase(parsedArgs, currentPhase)) {
Job pairwiseSimilarity =
prepareJob(
weightsPath,
pairwiseSimilarityPath,
CooccurrencesMapper.class,
IntWritable.class,
VectorWritable.class,
SimilarityReducer.class,
IntWritable.class,
VectorWritable.class);
pairwiseSimilarity.setCombinerClass(VectorSumReducer.class);
Configuration pairwiseConf = pairwiseSimilarity.getConfiguration();
pairwiseConf.set(THRESHOLD, String.valueOf(threshold));
pairwiseConf.set(NORMS_PATH, normsPath.toString());
pairwiseConf.set(NUM_NON_ZERO_ENTRIES_PATH, numNonZeroEntriesPath.toString());
pairwiseConf.set(MAXVALUES_PATH, maxValuesPath.toString());
pairwiseConf.set(SIMILARITY_CLASSNAME, similarityClassname);
pairwiseConf.setInt(NUMBER_OF_COLUMNS, numberOfColumns);
pairwiseConf.setBoolean(EXCLUDE_SELF_SIMILARITY, excludeSelfSimilarity);
boolean succeeded = pairwiseSimilarity.waitForCompletion(true);
if (!succeeded) {
return -1;
}
}
if (shouldRunNextPhase(parsedArgs, currentPhase)) {
Job asMatrix =
prepareJob(
pairwiseSimilarityPath,
getOutputPath(),
UnsymmetrifyMapper.class,
IntWritable.class,
VectorWritable.class,
MergeToTopKSimilaritiesReducer.class,
IntWritable.class,
VectorWritable.class);
asMatrix.setCombinerClass(MergeToTopKSimilaritiesReducer.class);
asMatrix.getConfiguration().setInt(MAX_SIMILARITIES_PER_ROW, maxSimilaritiesPerRow);
boolean succeeded = asMatrix.waitForCompletion(true);
if (!succeeded) {
return -1;
}
}
return 0;
}
