Many resources are needed to download a project. Please understand that we have to compensate our server costs. Thank you in advance. Project price only 1 $
You can buy this project and download/modify it how often you want.
/*******************************************************************************
* Copyright (c) 2015-2018 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.lossfunctions;
import org.nd4j.linalg.activations.IActivation;
import org.nd4j.linalg.api.ndarray.INDArray;
import org.nd4j.common.primitives.Pair;
import org.nd4j.serde.json.LegacyILossFunctionDeserializerHelper;
import org.nd4j.shade.jackson.annotation.JsonTypeInfo;
import java.io.Serializable;
/**
* Interface for loss functions
*/
@JsonTypeInfo(use = JsonTypeInfo.Id.CLASS, include = JsonTypeInfo.As.PROPERTY, property = "@class",
defaultImpl = LegacyILossFunctionDeserializerHelper.class)
public interface ILossFunction extends Serializable {
/**
* Compute the score (loss function value) for the given inputs.
* @param labels Label/expected preOutput
* @param preOutput Output of the model (neural network)
* @param activationFn Activation function that should be applied to preOutput
* @param mask Mask array; may be null
* @param average Whether the score should be averaged (divided by number of rows in labels/preOutput) or not @return Loss function value
*/
double computeScore(INDArray labels, INDArray preOutput, IActivation activationFn, INDArray mask, boolean average);
/**
* Compute the score (loss function value) for each example individually.
* For input [numExamples,nOut] returns scores as a column vector: [numExamples,1]
* @param labels Labels/expected output
* @param preOutput Output of the model (neural network)
* @param activationFn Activation function that should be applied to preOutput
* @param mask @return Loss function value for each example; column vector
*/
INDArray computeScoreArray(INDArray labels, INDArray preOutput, IActivation activationFn, INDArray mask);
/**
* Compute the gradient of the loss function with respect to the inputs: dL/dOutput
*
* @param labels Label/expected output
* @param preOutput Output of the model (neural network), before the activation function is applied
* @param activationFn Activation function that should be applied to preOutput
* @param mask Mask array; may be null
* @return Gradient dL/dPreOut
*/
INDArray computeGradient(INDArray labels, INDArray preOutput, IActivation activationFn, INDArray mask);
/**
* Compute both the score (loss function value) and gradient. This is equivalent to calling {@link #computeScore(INDArray, INDArray, IActivation, INDArray, boolean)}
* and {@link #computeGradient(INDArray, INDArray, IActivation, INDArray)} individually
*
* @param labels Label/expected output
* @param preOutput Output of the model (neural network)
* @param activationFn Activation function that should be applied to preOutput
* @param mask Mask array; may be null
* @param average Whether the score should be averaged (divided by number of rows in labels/output) or not
* @return The score (loss function value) and gradient
*/
//TODO: do we want to use the apache commons pair here?
Pair computeGradientAndScore(INDArray labels, INDArray preOutput, IActivation activationFn,
INDArray mask, boolean average);
/**
* The opName of this function
* @return
*/
String name();
}
