org.nd4j.linalg.lossfunctions.impl.LossSquaredHinge Maven / Gradle / Ivy
The newest version!
/*
* ******************************************************************************
* *
* *
* * This program and the accompanying materials are made available under the
* * terms of the Apache License, Version 2.0 which is available at
* * https://www.apache.org/licenses/LICENSE-2.0.
* *
* * See the NOTICE file distributed with this work for additional
* * information regarding copyright ownership.
* * 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.
* *
* * SPDX-License-Identifier: Apache-2.0
* *****************************************************************************
*/
package org.nd4j.linalg.lossfunctions.impl;
import lombok.EqualsAndHashCode;
import org.nd4j.common.base.Preconditions;
import org.nd4j.linalg.activations.IActivation;
import org.nd4j.linalg.api.ndarray.INDArray;
import org.nd4j.linalg.indexing.BooleanIndexing;
import org.nd4j.linalg.indexing.conditions.Conditions;
import org.nd4j.linalg.lossfunctions.ILossFunction;
import org.nd4j.linalg.lossfunctions.LossUtil;
import org.nd4j.common.primitives.Pair;
@EqualsAndHashCode
public class LossSquaredHinge implements ILossFunction {
public INDArray scoreArray(INDArray labels, INDArray preOutput, IActivation activationFn, INDArray mask) {
if(!labels.equalShapes(preOutput)){
Preconditions.throwEx("Labels and preOutput must have equal shapes: got shapes %s vs %s", labels.shape(), preOutput.shape());
}
labels = labels.castTo(preOutput.dataType()); //No-op if already correct dtype
/* y_hat is -1 or 1
hinge loss is max(0,1-y_hat*y)
*/
INDArray output = activationFn.getActivation(preOutput.dup(), true);
INDArray scoreArr = output.muli(labels); //y*yhat
scoreArr.rsubi(1.0); //1 - y*yhat
if (mask != null) {
LossUtil.applyMask(scoreArr, mask);
}
return scoreArr; // 1 - y*yhat
}
@Override
public double computeScore(INDArray labels, INDArray preOutput, IActivation activationFn, INDArray mask,
boolean average) {
INDArray scoreArr = computeScoreArray(labels, preOutput, activationFn, mask);
double score = scoreArr.sumNumber().doubleValue();
if (average)
score /= scoreArr.size(0);
return score;
}
@Override
public INDArray computeScoreArray(INDArray labels, INDArray preOutput, IActivation activationFn, INDArray mask) {
INDArray scoreArr = scoreArray(labels, preOutput, activationFn, mask);
BooleanIndexing.replaceWhere(scoreArr, 0.0, Conditions.lessThan(0.0));//max(0,1-y*yhat)
scoreArr.muli(scoreArr);
return scoreArr.sum(true,1);
}
@Override
public INDArray computeGradient(INDArray labels, INDArray preOutput, IActivation activationFn, INDArray mask) {
if(!labels.equalShapes(preOutput)){
Preconditions.throwEx("Labels and preOutput must have equal shapes: got shapes %s vs %s", labels.shape(), preOutput.shape());
}
labels = labels.castTo(preOutput.dataType()); //No-op if already correct dtype
INDArray scoreArr = scoreArray(labels, preOutput, activationFn, mask);
INDArray bitMaskRowCol = scoreArr.dup();
/*
bit mask is 0 if 1-sigma(y*yhat) is neg, bit mask is 1 if 1-sigma(y*yhat) is +ve
*/
BooleanIndexing.replaceWhere(bitMaskRowCol, 0.0, Conditions.lessThan(0.0));
BooleanIndexing.replaceWhere(bitMaskRowCol, 1.0, Conditions.greaterThan(0.0));
INDArray dLda = scoreArr.muli(2).muli(labels.neg());
dLda.muli(bitMaskRowCol);
if (mask != null && LossUtil.isPerOutputMasking(dLda, mask)) {
//For *most* activation functions: we don't actually need to mask dL/da in addition to masking dL/dz later
//but: some, like softmax, require both (due to dL/dz_i being a function of dL/da_j, for i != j)
//We could add a special case for softmax (activationFn instanceof ActivationSoftmax) but that would be
// error prone - though buy us a tiny bit of performance
LossUtil.applyMask(dLda, mask);
}
INDArray gradients = activationFn.backprop(preOutput, dLda).getFirst(); //TODO activation functions with params
if (mask != null) {
LossUtil.applyMask(gradients, mask);
}
return gradients;
}
@Override
public Pair computeGradientAndScore(INDArray labels,
INDArray preOutput, IActivation activationFn, INDArray mask, boolean average) {
//TODO: probably a more efficient way to do this...
return new Pair<>(computeScore(labels, preOutput, activationFn, mask, average),
computeGradient(labels, preOutput, activationFn, mask));
}
/**
* The opName of this function
*
* @return
*/
@Override
public String name() {
return toString();
}
@Override
public String toString() {
return "LossSquaredHinge()";
}
}