OnlineHybridTrainer.java

package edu.fudan.ml.classifier.struct;

import java.io.IOException;
import java.util.Arrays;
import java.util.List;

import edu.fudan.ml.classifier.linear.Linear;
import edu.fudan.ml.classifier.linear.OnlineTrainer;
import edu.fudan.ml.classifier.linear.inf.Inferencer;
import edu.fudan.ml.classifier.linear.update.Update;
import edu.fudan.ml.loss.Loss;
import edu.fudan.ml.types.Instance;
import edu.fudan.ml.types.InstanceSet;
import edu.fudan.util.MyArrays;

public class OnlineHybridTrainer extends OnlineTrainer {

	public OnlineHybridTrainer(Inferencer msolver, Update update,
			Loss loss, int fsize, int iternum, float c) {
		super(msolver, update, loss, fsize, iternum, c);
	}

	public Linear train(InstanceSet trainset, InstanceSet devset) {
		int numSamples = trainset.size();
		int count = 0;
		for (int ii = 0; ii < numSamples; ii++) {
			Instance inst = trainset.getInstance(ii);
			int[][] targets = (int[][]) inst.getTarget();
			for(int i = 0; i < targets.length; i++)	{
				count += targets[i].length;
			}
		}

		System.out.println("Training Number: "+numSamples);
		System.out.println("Chars Number: " + count);

		float hisErrRate = Float.MAX_VALUE;

		long beginTime, endTime;
		long beginTimeIter, endTimeIter;
		int iter = 0;
		int frac = numSamples / 10;

		float[] averageWeights = null;
		if (method == TrainMethod.Average || method == TrainMethod.FastAverage) {
			averageWeights = new float[weights.length];
		}

		beginTime = System.currentTimeMillis();

		if (shuffle)
			trainset.shuffle(random);

		while (iter++ < iternum) {
			if (!simpleOutput) {
				System.out.print("iter:");
				System.out.print(iter + "\t");
			}
			float err = 0;
			float errtot = 0;
			int progress = frac;

			beginTimeIter = System.currentTimeMillis();

			float[] innerWeights = null;
			if (method == TrainMethod.Average) {
				innerWeights = Arrays.copyOf(weights, weights.length);
			}

			for (int ii = 0; ii < numSamples; ii++) {
				Instance inst = trainset.getInstance(ii);
				List pred = (List) inferencer.getBest(inst, 1);
				float l = loss.calc((int[][]) pred.get(0), (int[][]) inst.getTarget());
				if (l > 0)	{
					err += l;
					errtot++;
					update.update(inst, weights, pred.get(0), c);
				}
				if (method == TrainMethod.Average) {
					for (int i = 0; i < weights.length; i++) {
						innerWeights[i] += weights[i];
					}
				}

				if (DEBUG && l > 0) {
					pred = (List) inferencer.getBest(inst, 1);
					l = loss.calc((int[]) pred.get(0), (int[]) inst.getTarget());
				}

				if (!simpleOutput && ii % progress == 0) {
					System.out.print('.');
					progress += frac;
				}
			}

			float curErrRate = err / count;

			endTimeIter = System.currentTimeMillis();

			if (!simpleOutput) {
				System.out.println("\ttime:" + (endTimeIter - beginTimeIter)
						/ 1000.0 + "s");
				System.out.print("Train:");
				System.out.print("\tTag acc:");
			}
			System.out.print(1 - curErrRate);
			if (!simpleOutput) {
				System.out.print("\tSentence acc:");
				System.out.print(1 - errtot / numSamples);
				System.out.println();
			}
			
			System.out.print("Weight Numbers: "+MyArrays.countNoneZero(weights));
			if (innerOptimized) {
				int[] idx = MyArrays.getTop(weights.clone(), threshold, false);
				MyArrays.set(weights, idx, 0.0f);
				System.out.print("	After Optimized: " + MyArrays.countNoneZero(weights));
			}
			System.out.println();
			
			if (devset != null) {
				evaluate(devset);
			}
			
			if (method == TrainMethod.Average) {
				for (int i = 0; i < innerWeights.length; i++) {
					averageWeights[i] += innerWeights[i] / numSamples;
				}
			} else if (method == TrainMethod.FastAverage) {
				for (int i = 0; i < weights.length; i++) {
					averageWeights[i] += weights[i];
				}
			}
			
			if (interim) {
				Linear p = new Linear(inferencer,
						trainset.getAlphabetFactory());
				try {
					p.saveTo("tmp.model");
				} catch (IOException e) {
					System.err.println("write model error!");
				}
			}

		}

		if (method == TrainMethod.Average || method == TrainMethod.FastAverage) {
			for (int i = 0; i < averageWeights.length; i++) {
				averageWeights[i] /= iternum;
			}
			weights = null;
			weights = averageWeights;
			inferencer.setWeights(weights);
		}

		System.out.print("Weight Numbers: "+MyArrays.countNoneZero(weights));
		if (finalOptimized) {
			int[] idx = MyArrays.getTop(weights.clone(), threshold, false);
			MyArrays.set(weights, idx, 0.0f);
			System.out.print("	After Optimized: " + MyArrays.countNoneZero(weights));
		}
		System.out.println();

		endTime = System.currentTimeMillis();
		System.out.println("time escape:" + (endTime - beginTime) / 1000.0
				+ "s");
		Linear p = new Linear(inferencer, trainset.getAlphabetFactory());
		return p;
	}

	public void evaluate(InstanceSet devset) {
		float err = 0;
		float errtot = 0;
		int total = 0;
		for (int i = 0; i < devset.size(); i++) {
			Instance inst = devset.getInstance(i);
			total += ((int[]) inst.getTarget()).length;
			List pred = (List) inferencer.getBest(inst, 1);
			float l = loss.calc(pred.get(0), inst.getTarget());
			if (l > 0) {
				errtot += 1.0;
				err += l;
			}

		}
		if (!simpleOutput) {
			System.out.print("Test:\t");
			System.out.print(total - err);
			System.out.print('/');
			System.out.print(total);
			System.out.print("\tTag acc:");
		} else {
			System.out.print('\t');
		}
		System.out.print(1 - err / total);
		if (!simpleOutput) {
			System.out.print("\tSentence acc:");
			System.out.println(1 - errtot / devset.size());

		}
		System.out.println();
	}
}