org.nd4j.linalg.api.ops.impl.accum.Variance Maven / Gradle / Ivy

Show more of this group Show more artifacts with this name
Show all versions of nd4j-api Show documentation
There is a newer version: 1.0.0-M2.1
/*******************************************************************************
 * 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.api.ops.impl.accum;

import lombok.val;
import org.nd4j.autodiff.samediff.SDVariable;
import org.nd4j.autodiff.samediff.SameDiff;
import org.nd4j.imports.NoOpNameFoundException;
import org.nd4j.linalg.api.ndarray.INDArray;
import org.nd4j.linalg.api.ops.BaseAccumulation;
import org.nd4j.linalg.api.ops.executioner.OpExecutioner;
import org.nd4j.linalg.api.shape.Shape;
import org.nd4j.linalg.factory.Nd4j;

import java.util.Arrays;
import java.util.Collections;
import java.util.List;

/**
 * Variance with bias correction.
 * Bias can either be divided by n or adjusted with:
 * (currentResult - (pow(bias, 2.0) / n())) / (n() - 1.0);
 *
 * @author Adam Gibson
 */
public class Variance extends BaseAccumulation {
    protected double mean, bias;
    protected boolean biasCorrected = true;

    public Variance(SameDiff sameDiff, SDVariable i_v, boolean biasCorrected, boolean keepDims, int[] dimensions) {
        super(sameDiff, i_v, dimensions, keepDims);
        this.biasCorrected = biasCorrected;
    }

    public Variance() {
    }

    public Variance(boolean biasCorrected) {
        this.biasCorrected = biasCorrected;
    }

    public Variance(INDArray x, INDArray y, INDArray z, long n) {
        super(x, y, z, n);
        init(x, y, z, n);
    }

    public Variance(INDArray x, INDArray y, long n) {
        this(x, y, x, n);
    }

    public Variance(INDArray x) {
        this(x, null, x, x.lengthLong(), true);
    }

    public Variance(INDArray x, INDArray y) {
        super(x, y);
    }

    public Variance(INDArray x, INDArray y, INDArray z, long n, boolean biasCorrected) {
        super(x, y, z, n);
        this.biasCorrected = biasCorrected;
        init(x, y, z, n);
    }

    public Variance(INDArray x, INDArray y, long n, boolean biasCorrected) {
        super(x, y, n);
        this.biasCorrected = biasCorrected;
        init(x, y, z, n);
    }

    public Variance(INDArray x, boolean biasCorrected) {
        super(x);
        this.biasCorrected = biasCorrected;
        init(x, y, z, n);
    }

    public Variance(INDArray x, INDArray y, boolean biasCorrected) {
        super(x, y);
        this.biasCorrected = biasCorrected;
        init(x, y, x, x.lengthLong());
    }

    public Variance(INDArray x, INDArray y, INDArray z, boolean newFormat, boolean keepDims, int[] dimensions) {
        super(x, y, z, newFormat, keepDims, dimensions);
    }

    @Override
    public INDArray noOp() {
        return Nd4j.zerosLike(x());
    }

    @Override
    public int opNum() {
        return 0;
    }

    @Override
    public String opName() {
        return "var";
    }


    @Override
    public void init(INDArray x, INDArray y, INDArray z, long n) {
        super.init(x, y, z, n);
        if (Nd4j.executionMode == OpExecutioner.ExecutionMode.JAVA) {
            if (biasCorrected)
                this.bias = Nd4j.getExecutioner().execAndReturn(new Bias(x)).getFinalResult().doubleValue();
            mean = Nd4j.getExecutioner().execAndReturn(new Mean(x)).getFinalResult().doubleValue();
        }

    }

    @Override
    public boolean isPassThrough() {
        return true;
    }


    public boolean isBiasCorrected() {
        return biasCorrected;
    }

    public void setBiasCorrected(boolean biasCorrected) {
        this.biasCorrected = biasCorrected;
    }


    @Override
    public List doDiff(List grad) {
        //If out = var(in) then:
        //dL/dIn = dL/dOut * dOut/dIn
        // with dOut/dIn = (in-mean) * 2/(n-1)
        return Collections.singletonList(f().varianceBp(arg(), grad.get(0), biasCorrected, keepDims, dimensions));
    }

    @Override
    public String onnxName() {
        throw new NoOpNameFoundException("No onnx opName found for " + opName());
    }

    @Override
    public String tensorflowName() {
        return "moments";
    }


    @Override
    public Type getOpType() {
        return Type.VARIANCE;
    }
}