Examples with SimpleMatrix org.ejml.simple.SimpleMatrix used on opensource projects

Example 56 with SimpleMatrix

use of org.ejml.simple.SimpleMatrix in project CoreNLP by stanfordnlp.

the class DVParserCostAndGradient method backpropDerivative.

public void backpropDerivative(Tree tree, List<String> words, IdentityHashMap<Tree, SimpleMatrix> nodeVectors, TwoDimensionalMap<String, String, SimpleMatrix> binaryW_dfs, Map<String, SimpleMatrix> unaryW_dfs, TwoDimensionalMap<String, String, SimpleMatrix> binaryScoreDerivatives, Map<String, SimpleMatrix> unaryScoreDerivatives, Map<String, SimpleMatrix> wordVectorDerivatives) {
    SimpleMatrix delta = new SimpleMatrix(op.lexOptions.numHid, 1);
    backpropDerivative(tree, words, nodeVectors, binaryW_dfs, unaryW_dfs, binaryScoreDerivatives, unaryScoreDerivatives, wordVectorDerivatives, delta);
}

Example 57 with SimpleMatrix

use of org.ejml.simple.SimpleMatrix in project CoreNLP by stanfordnlp.

the class DumpMatrices method main.

public static void main(String[] args) throws IOException {
    String modelPath = null;
    String outputDir = null;
    for (int argIndex = 0; argIndex < args.length; ) {
        if (args[argIndex].equalsIgnoreCase("-model")) {
            modelPath = args[argIndex + 1];
            argIndex += 2;
        } else if (args[argIndex].equalsIgnoreCase("-output")) {
            outputDir = args[argIndex + 1];
            argIndex += 2;
        } else {
            log.info("Unknown argument " + args[argIndex]);
            help();
        }
    }
    if (outputDir == null || modelPath == null) {
        help();
    }
    File outputFile = new File(outputDir);
    FileSystem.checkNotExistsOrFail(outputFile);
    FileSystem.mkdirOrFail(outputFile);
    LexicalizedParser parser = LexicalizedParser.loadModel(modelPath);
    DVModel model = DVParser.getModelFromLexicalizedParser(parser);
    String binaryWDir = outputDir + File.separator + "binaryW";
    FileSystem.mkdirOrFail(binaryWDir);
    for (TwoDimensionalMap.Entry<String, String, SimpleMatrix> entry : model.binaryTransform) {
        String filename = binaryWDir + File.separator + entry.getFirstKey() + "_" + entry.getSecondKey() + ".txt";
        dumpMatrix(filename, entry.getValue());
    }
    String binaryScoreDir = outputDir + File.separator + "binaryScore";
    FileSystem.mkdirOrFail(binaryScoreDir);
    for (TwoDimensionalMap.Entry<String, String, SimpleMatrix> entry : model.binaryScore) {
        String filename = binaryScoreDir + File.separator + entry.getFirstKey() + "_" + entry.getSecondKey() + ".txt";
        dumpMatrix(filename, entry.getValue());
    }
    String unaryWDir = outputDir + File.separator + "unaryW";
    FileSystem.mkdirOrFail(unaryWDir);
    for (Map.Entry<String, SimpleMatrix> entry : model.unaryTransform.entrySet()) {
        String filename = unaryWDir + File.separator + entry.getKey() + ".txt";
        dumpMatrix(filename, entry.getValue());
    }
    String unaryScoreDir = outputDir + File.separator + "unaryScore";
    FileSystem.mkdirOrFail(unaryScoreDir);
    for (Map.Entry<String, SimpleMatrix> entry : model.unaryScore.entrySet()) {
        String filename = unaryScoreDir + File.separator + entry.getKey() + ".txt";
        dumpMatrix(filename, entry.getValue());
    }
    String embeddingFile = outputDir + File.separator + "embeddings.txt";
    FileWriter fout = new FileWriter(embeddingFile);
    BufferedWriter bout = new BufferedWriter(fout);
    for (Map.Entry<String, SimpleMatrix> entry : model.wordVectors.entrySet()) {
        bout.write(entry.getKey());
        SimpleMatrix vector = entry.getValue();
        for (int i = 0; i < vector.numRows(); ++i) {
            bout.write("  " + vector.get(i, 0));
        }
        bout.write("\n");
    }
    bout.close();
    fout.close();
}

Example 58 with SimpleMatrix

use of org.ejml.simple.SimpleMatrix in project CoreNLP by stanfordnlp.

the class FindNearestNeighbors method main.

public static void main(String[] args) throws Exception {
    String modelPath = null;
    String outputPath = null;
    String testTreebankPath = null;
    FileFilter testTreebankFilter = null;
    List<String> unusedArgs = new ArrayList<>();
    for (int argIndex = 0; argIndex < args.length; ) {
        if (args[argIndex].equalsIgnoreCase("-model")) {
            modelPath = args[argIndex + 1];
            argIndex += 2;
        } else if (args[argIndex].equalsIgnoreCase("-testTreebank")) {
            Pair<String, FileFilter> treebankDescription = ArgUtils.getTreebankDescription(args, argIndex, "-testTreebank");
            argIndex = argIndex + ArgUtils.numSubArgs(args, argIndex) + 1;
            testTreebankPath = treebankDescription.first();
            testTreebankFilter = treebankDescription.second();
        } else if (args[argIndex].equalsIgnoreCase("-output")) {
            outputPath = args[argIndex + 1];
            argIndex += 2;
        } else {
            unusedArgs.add(args[argIndex++]);
        }
    }
    if (modelPath == null) {
        throw new IllegalArgumentException("Need to specify -model");
    }
    if (testTreebankPath == null) {
        throw new IllegalArgumentException("Need to specify -testTreebank");
    }
    if (outputPath == null) {
        throw new IllegalArgumentException("Need to specify -output");
    }
    String[] newArgs = unusedArgs.toArray(new String[unusedArgs.size()]);
    LexicalizedParser lexparser = LexicalizedParser.loadModel(modelPath, newArgs);
    Treebank testTreebank = null;
    if (testTreebankPath != null) {
        log.info("Reading in trees from " + testTreebankPath);
        if (testTreebankFilter != null) {
            log.info("Filtering on " + testTreebankFilter);
        }
        testTreebank = lexparser.getOp().tlpParams.memoryTreebank();
        ;
        testTreebank.loadPath(testTreebankPath, testTreebankFilter);
        log.info("Read in " + testTreebank.size() + " trees for testing");
    }
    FileWriter out = new FileWriter(outputPath);
    BufferedWriter bout = new BufferedWriter(out);
    log.info("Parsing " + testTreebank.size() + " trees");
    int count = 0;
    List<ParseRecord> records = Generics.newArrayList();
    for (Tree goldTree : testTreebank) {
        List<Word> tokens = goldTree.yieldWords();
        ParserQuery parserQuery = lexparser.parserQuery();
        if (!parserQuery.parse(tokens)) {
            throw new AssertionError("Could not parse: " + tokens);
        }
        if (!(parserQuery instanceof RerankingParserQuery)) {
            throw new IllegalArgumentException("Expected a LexicalizedParser with a Reranker attached");
        }
        RerankingParserQuery rpq = (RerankingParserQuery) parserQuery;
        if (!(rpq.rerankerQuery() instanceof DVModelReranker.Query)) {
            throw new IllegalArgumentException("Expected a LexicalizedParser with a DVModel attached");
        }
        DeepTree tree = ((DVModelReranker.Query) rpq.rerankerQuery()).getDeepTrees().get(0);
        SimpleMatrix rootVector = null;
        for (Map.Entry<Tree, SimpleMatrix> entry : tree.getVectors().entrySet()) {
            if (entry.getKey().label().value().equals("ROOT")) {
                rootVector = entry.getValue();
                break;
            }
        }
        if (rootVector == null) {
            throw new AssertionError("Could not find root nodevector");
        }
        out.write(tokens + "\n");
        out.write(tree.getTree() + "\n");
        for (int i = 0; i < rootVector.getNumElements(); ++i) {
            out.write("  " + rootVector.get(i));
        }
        out.write("\n\n\n");
        count++;
        if (count % 10 == 0) {
            log.info("  " + count);
        }
        records.add(new ParseRecord(tokens, goldTree, tree.getTree(), rootVector, tree.getVectors()));
    }
    log.info("  done parsing");
    List<Pair<Tree, SimpleMatrix>> subtrees = Generics.newArrayList();
    for (ParseRecord record : records) {
        for (Map.Entry<Tree, SimpleMatrix> entry : record.nodeVectors.entrySet()) {
            if (entry.getKey().getLeaves().size() <= maxLength) {
                subtrees.add(Pair.makePair(entry.getKey(), entry.getValue()));
            }
        }
    }
    log.info("There are " + subtrees.size() + " subtrees in the set of trees");
    PriorityQueue<ScoredObject<Pair<Tree, Tree>>> bestmatches = new PriorityQueue<>(101, ScoredComparator.DESCENDING_COMPARATOR);
    for (int i = 0; i < subtrees.size(); ++i) {
        log.info(subtrees.get(i).first().yieldWords());
        log.info(subtrees.get(i).first());
        for (int j = 0; j < subtrees.size(); ++j) {
            if (i == j) {
                continue;
            }
            // TODO: look at basic category?
            double normF = subtrees.get(i).second().minus(subtrees.get(j).second()).normF();
            bestmatches.add(new ScoredObject<>(Pair.makePair(subtrees.get(i).first(), subtrees.get(j).first()), normF));
            if (bestmatches.size() > 100) {
                bestmatches.poll();
            }
        }
        List<ScoredObject<Pair<Tree, Tree>>> ordered = Generics.newArrayList();
        while (bestmatches.size() > 0) {
            ordered.add(bestmatches.poll());
        }
        Collections.reverse(ordered);
        for (ScoredObject<Pair<Tree, Tree>> pair : ordered) {
            log.info(" MATCHED " + pair.object().second.yieldWords() + " ... " + pair.object().second() + " with a score of " + pair.score());
        }
        log.info();
        log.info();
        bestmatches.clear();
    }
    /*
    for (int i = 0; i < records.size(); ++i) {
      if (i % 10 == 0) {
        log.info("  " + i);
      }
      List<ScoredObject<ParseRecord>> scored = Generics.newArrayList();
      for (int j = 0; j < records.size(); ++j) {
        if (i == j) continue;

        double score = 0.0;
        int matches = 0;
        for (Map.Entry<Tree, SimpleMatrix> first : records.get(i).nodeVectors.entrySet()) {
          for (Map.Entry<Tree, SimpleMatrix> second : records.get(j).nodeVectors.entrySet()) {
            String firstBasic = dvparser.dvModel.basicCategory(first.getKey().label().value());
            String secondBasic = dvparser.dvModel.basicCategory(second.getKey().label().value());
            if (firstBasic.equals(secondBasic)) {
              ++matches;
              double normF = first.getValue().minus(second.getValue()).normF();
              score += normF * normF;
            }
          }
        }
        if (matches == 0) {
          score = Double.POSITIVE_INFINITY;
        } else {
          score = score / matches;
        }
        //double score = records.get(i).vector.minus(records.get(j).vector).normF();
        scored.add(new ScoredObject<ParseRecord>(records.get(j), score));
      }
      Collections.sort(scored, ScoredComparator.ASCENDING_COMPARATOR);

      out.write(records.get(i).sentence.toString() + "\n");
      for (int j = 0; j < numNeighbors; ++j) {
        out.write("   " + scored.get(j).score() + ": " + scored.get(j).object().sentence + "\n");
      }
      out.write("\n\n");
    }
    log.info();
    */
    bout.flush();
    out.flush();
    out.close();
}

Example 59 with SimpleMatrix

use of org.ejml.simple.SimpleMatrix in project CoreNLP by stanfordnlp.

the class ParseAndPrintMatrices method main.

public static void main(String[] args) throws IOException {
    String modelPath = null;
    String outputPath = null;
    String inputPath = null;
    String testTreebankPath = null;
    FileFilter testTreebankFilter = null;
    List<String> unusedArgs = Generics.newArrayList();
    for (int argIndex = 0; argIndex < args.length; ) {
        if (args[argIndex].equalsIgnoreCase("-model")) {
            modelPath = args[argIndex + 1];
            argIndex += 2;
        } else if (args[argIndex].equalsIgnoreCase("-output")) {
            outputPath = args[argIndex + 1];
            argIndex += 2;
        } else if (args[argIndex].equalsIgnoreCase("-input")) {
            inputPath = args[argIndex + 1];
            argIndex += 2;
        } else if (args[argIndex].equalsIgnoreCase("-testTreebank")) {
            Pair<String, FileFilter> treebankDescription = ArgUtils.getTreebankDescription(args, argIndex, "-testTreebank");
            argIndex = argIndex + ArgUtils.numSubArgs(args, argIndex) + 1;
            testTreebankPath = treebankDescription.first();
            testTreebankFilter = treebankDescription.second();
        } else {
            unusedArgs.add(args[argIndex++]);
        }
    }
    String[] newArgs = unusedArgs.toArray(new String[unusedArgs.size()]);
    LexicalizedParser parser = LexicalizedParser.loadModel(modelPath, newArgs);
    DVModel model = DVParser.getModelFromLexicalizedParser(parser);
    File outputFile = new File(outputPath);
    FileSystem.checkNotExistsOrFail(outputFile);
    FileSystem.mkdirOrFail(outputFile);
    int count = 0;
    if (inputPath != null) {
        Reader input = new BufferedReader(new FileReader(inputPath));
        DocumentPreprocessor processor = new DocumentPreprocessor(input);
        for (List<HasWord> sentence : processor) {
            // index from 1
            count++;
            ParserQuery pq = parser.parserQuery();
            if (!(pq instanceof RerankingParserQuery)) {
                throw new IllegalArgumentException("Expected a RerankingParserQuery");
            }
            RerankingParserQuery rpq = (RerankingParserQuery) pq;
            if (!rpq.parse(sentence)) {
                throw new RuntimeException("Unparsable sentence: " + sentence);
            }
            RerankerQuery reranker = rpq.rerankerQuery();
            if (!(reranker instanceof DVModelReranker.Query)) {
                throw new IllegalArgumentException("Expected a DVModelReranker");
            }
            DeepTree deepTree = ((DVModelReranker.Query) reranker).getDeepTrees().get(0);
            IdentityHashMap<Tree, SimpleMatrix> vectors = deepTree.getVectors();
            for (Map.Entry<Tree, SimpleMatrix> entry : vectors.entrySet()) {
                log.info(entry.getKey() + "   " + entry.getValue());
            }
            FileWriter fout = new FileWriter(outputPath + File.separator + "sentence" + count + ".txt");
            BufferedWriter bout = new BufferedWriter(fout);
            bout.write(SentenceUtils.listToString(sentence));
            bout.newLine();
            bout.write(deepTree.getTree().toString());
            bout.newLine();
            for (HasWord word : sentence) {
                outputMatrix(bout, model.getWordVector(word.word()));
            }
            Tree rootTree = findRootTree(vectors);
            outputTreeMatrices(bout, rootTree, vectors);
            bout.flush();
            fout.close();
        }
    }
}

Example 60 with SimpleMatrix

use of org.ejml.simple.SimpleMatrix in project CoreNLP by stanfordnlp.

the class AverageDVModels method averageUnaryMatrices.

public static Map<String, SimpleMatrix> averageUnaryMatrices(List<Map<String, SimpleMatrix>> maps) {
    Map<String, SimpleMatrix> averages = Generics.newTreeMap();
    for (String name : getUnaryMatrixNames(maps)) {
        int count = 0;
        SimpleMatrix matrix = null;
        for (Map<String, SimpleMatrix> map : maps) {
            if (!map.containsKey(name)) {
                continue;
            }
            SimpleMatrix original = map.get(name);
            ++count;
            if (matrix == null) {
                matrix = original;
            } else {
                matrix = matrix.plus(original);
            }
        }
        matrix = matrix.divide(count);
        averages.put(name, matrix);
    }
    return averages;
}