org.nd4j.linalg.api.ops.impl.loss.HuberLoss Maven / Gradle / Ivy
/*******************************************************************************
* Copyright (c) 2015-2019 Skymind, Inc.
*
* 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.
*
* 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.api.ops.impl.loss;
import org.nd4j.autodiff.loss.LossReduce;
import org.nd4j.autodiff.samediff.SDVariable;
import org.nd4j.autodiff.samediff.SameDiff;
import org.nd4j.base.Preconditions;
import java.util.Arrays;
import java.util.List;
/**
* Huber loss
*
* @author Alex Black
*/
public class HuberLoss extends BaseLoss {
private double delta;
public HuberLoss(SameDiff sameDiff, LossReduce lossReduce, SDVariable predictions, SDVariable weights, SDVariable labels, double delta){
super(sameDiff, lossReduce, predictions, weights, labels);
Preconditions.checkState(delta >= 0.0, "Delta must be >= 0.0. Got: %s", delta);
this.delta = delta;
tArguments.add(delta);
}
public HuberLoss(){ }
@Override
public String opName() {
return "huber_loss";
}
@Override
public List doDiff(List grad){
//No external gradient
//Args are: predictions, weights, label
SDVariable[] grads = f().lossHuberBp(arg(2), arg(0), arg(1), lossReduce, delta);
return Arrays.asList(grads);
}
}