org.nd4j.linalg.ops.transforms.Transforms Maven / Gradle / Ivy
/*******************************************************************************
* 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.ops.transforms;
import lombok.NonNull;
import lombok.val;
import org.nd4j.base.Preconditions;
import org.nd4j.linalg.api.buffer.DataType;
import org.nd4j.linalg.api.ndarray.INDArray;
import org.nd4j.linalg.api.ops.CustomOp;
import org.nd4j.linalg.api.ops.ScalarOp;
import org.nd4j.linalg.api.ops.TransformOp;
import org.nd4j.linalg.api.ops.impl.reduce3.*;
import org.nd4j.linalg.api.ops.impl.scalar.*;
import org.nd4j.linalg.api.ops.impl.scalar.comparison.ScalarNot;
import org.nd4j.linalg.api.ops.impl.shape.Cross;
import org.nd4j.linalg.api.ops.impl.transforms.bool.BooleanNot;
import org.nd4j.linalg.api.ops.impl.transforms.any.IsMax;
import org.nd4j.linalg.api.ops.impl.transforms.custom.ATan2;
import org.nd4j.linalg.api.ops.impl.transforms.custom.GreaterThanOrEqual;
import org.nd4j.linalg.api.ops.impl.transforms.custom.LessThanOrEqual;
import org.nd4j.linalg.api.ops.impl.transforms.custom.Reverse;
import org.nd4j.linalg.api.ops.impl.transforms.custom.SoftMax;
import org.nd4j.linalg.api.ops.impl.transforms.floating.*;
import org.nd4j.linalg.api.ops.impl.transforms.comparison.*;
import org.nd4j.linalg.api.ops.impl.transforms.gradient.EluBp;
import org.nd4j.linalg.api.ops.impl.transforms.gradient.HardTanhDerivative;
import org.nd4j.linalg.api.ops.impl.transforms.gradient.LeakyReLUDerivative;
import org.nd4j.linalg.api.ops.impl.transforms.gradient.SoftSignDerivative;
import org.nd4j.linalg.api.ops.impl.transforms.pairwise.arithmetic.PowPairwise;
import org.nd4j.linalg.api.ops.impl.transforms.pairwise.bool.And;
import org.nd4j.linalg.api.ops.impl.transforms.pairwise.bool.Or;
import org.nd4j.linalg.api.ops.impl.transforms.pairwise.bool.Xor;
import org.nd4j.linalg.api.ops.impl.transforms.same.*;
import org.nd4j.linalg.api.ops.impl.transforms.strict.*;
import org.nd4j.linalg.api.shape.LongShapeDescriptor;
import org.nd4j.linalg.api.shape.Shape;
import org.nd4j.linalg.factory.Nd4j;
import org.nd4j.linalg.inverse.InvertMatrix;
import java.util.Arrays;
import java.util.List;
/**
* Functional interface for the different op classes
*
* @author Adam Gibson
*/
public class Transforms {
private Transforms() {
}
/**
* Cosine similarity
*
* @param d1 the first vector
* @param d2 the second vector
* @return the cosine similarities between the 2 arrays
*/
public static double cosineSim(@NonNull INDArray d1, @NonNull INDArray d2) {
return Nd4j.getExecutioner().exec(new CosineSimilarity(d1, d2)).getDouble(0);
}
public static double cosineDistance(@NonNull INDArray d1, @NonNull INDArray d2) {
return Nd4j.getExecutioner().exec(new CosineDistance(d1, d2)).getDouble(0);
}
public static double hammingDistance(@NonNull INDArray d1, @NonNull INDArray d2) {
return Nd4j.getExecutioner().exec(new HammingDistance(d1, d2)).getDouble(0);
}
public static double jaccardDistance(@NonNull INDArray d1, @NonNull INDArray d2) {
return Nd4j.getExecutioner().exec(new JaccardDistance(d1, d2)).getDouble(0);
}
public static INDArray allCosineSimilarities(@NonNull INDArray d1, @NonNull INDArray d2, int... dimensions) {
return Nd4j.getExecutioner().exec(new CosineSimilarity(d1, d2, true, dimensions));
}
public static INDArray allCosineDistances(@NonNull INDArray d1, @NonNull INDArray d2, int... dimensions) {
return Nd4j.getExecutioner().exec(new CosineDistance(d1, d2, true, dimensions));
}
public static INDArray allEuclideanDistances(@NonNull INDArray d1, @NonNull INDArray d2, int... dimensions) {
return Nd4j.getExecutioner().exec(new EuclideanDistance(d1, d2, true, dimensions));
}
public static INDArray allManhattanDistances(@NonNull INDArray d1, @NonNull INDArray d2, int... dimensions) {
return Nd4j.getExecutioner().exec(new ManhattanDistance(d1, d2, true, dimensions));
}
public static INDArray reverse(INDArray x, boolean dup) {
return Nd4j.getExecutioner().exec(new Reverse(x, dup ? x.ulike() : x))[0];
}
/**
* Dot product, new INDArray instance will be returned.
* Note that the Nd4J design is different from Numpy. Numpy dot on 2d arrays is matrix multiplication. Nd4J is
* full array dot product reduction.
*
* @param x the first vector
* @param y the second vector
* @return the dot product between the 2 arrays
*/
public static INDArray dot(INDArray x, INDArray y) {
return Nd4j.getExecutioner().exec(new Dot(x, y));
}
public static INDArray cross(INDArray x, INDArray y) {
Cross c = new Cross(x, y, null);
List shape = c.calculateOutputShape();
INDArray out = Nd4j.create(shape.get(0));
c.addOutputArgument(out);
Nd4j.getExecutioner().exec(c);
return out;
}
/**
* @param d1
* @param d2
* @return
*/
public static double manhattanDistance(@NonNull INDArray d1, @NonNull INDArray d2) {
return d1.distance1(d2);
}
/**
* Atan2 operation, new INDArray instance will be returned
* Note the order of x and y parameters is opposite to that of {@link java.lang.Math#atan2(double, double)}
*
* @param x the abscissa coordinate
* @param y the ordinate coordinate
* @return the theta from point (r, theta) when converting (x,y) from to cartesian to polar coordinates
*/
public static INDArray atan2(@NonNull INDArray x, @NonNull INDArray y) {
// Switched on purpose, to match OldATan2 (which the javadoc was written for)
return Nd4j.getExecutioner().exec(new ATan2(y, x, x.ulike()))[0];
}
/**
* @param d1
* @param d2
* @return
*/
public static double euclideanDistance(@NonNull INDArray d1, @NonNull INDArray d2) {
return d1.distance2(d2);
}
/**
* Normalize data to zero mean and unit variance
* substract by the mean and divide by the standard deviation
*
* @param toNormalize the ndarray to normalize
* @return the normalized ndarray
*/
public static INDArray normalizeZeroMeanAndUnitVariance(INDArray toNormalize) {
INDArray columnMeans = toNormalize.mean(0);
INDArray columnStds = toNormalize.std(0);
toNormalize.subiRowVector(columnMeans);
//padding for non zero
columnStds.addi(Nd4j.EPS_THRESHOLD);
toNormalize.diviRowVector(columnStds);
return toNormalize;
}
/**
* Scale by 1 / norm2 of the matrix
*
* @param toScale the ndarray to scale
* @return the scaled ndarray
*/
public static INDArray unitVec(INDArray toScale) {
double length = toScale.norm2Number().doubleValue();
if (length > 0) {
if (toScale.data().dataType() == (DataType.FLOAT))
return Nd4j.getBlasWrapper().scal(1.0f / (float) length, toScale);
else
return Nd4j.getBlasWrapper().scal(1.0 / length, toScale);
}
return toScale;
}
/**
* Returns the negative of an ndarray
*
* @param ndArray the ndarray to take the negative of
* @return the negative of the ndarray
*/
public static INDArray neg(INDArray ndArray) {
return neg(ndArray, true);
}
/**
* Binary matrix of whether the number at a given index is greater than
*
* @param ndArray
* @return
*/
public static INDArray floor(INDArray ndArray) {
return floor(ndArray, true);
}
/**
* Binary matrix of whether the number at a given index is greater than
*
* @param ndArray
* @return
*/
public static INDArray ceiling(INDArray ndArray) {
return ceiling(ndArray, true);
}
/**
* Ceiling function
*
* @param ndArray
* @param copyOnOps
* @return
*/
public static INDArray ceiling(INDArray ndArray, boolean copyOnOps) {
return exec(copyOnOps ? new Ceil(ndArray, ndArray.ulike()) : new Ceil(ndArray, ndArray));
}
/**
* Signum function of this ndarray
*
* @param toSign
* @return
*/
public static INDArray sign(INDArray toSign) {
return sign(toSign, true);
}
/**
* @param ndArray
* @param k
* @return
*/
public static INDArray stabilize(INDArray ndArray, double k) {
return stabilize(ndArray, k, true);
}
/**
* Sin function
*
* @param in
* @return
*/
public static INDArray sin(INDArray in) {
return sin(in, true);
}
/**
* Sin function
*
* @param in
* @param copy
* @return
*/
public static INDArray sin(INDArray in, boolean copy) {
return Nd4j.getExecutioner().exec(new Sin(in, (copy ? in.ulike() : in)));
}
/**
* Sin function
*
* @param in
* @return
*/
public static INDArray atanh(INDArray in) {
return atanh(in, true);
}
/**
* Sin function
*
* @param in
* @param copy
* @return
*/
public static INDArray atanh(INDArray in, boolean copy) {
return Nd4j.getExecutioner().exec(new ATanh(in, (copy ? in.ulike() : in)));
}
/**
* Sinh function
*
* @param in
* @return
*/
public static INDArray sinh(INDArray in) {
return sinh(in, true);
}
/**
* Sinh function
*
* @param in
* @param copy
* @return
*/
public static INDArray sinh(INDArray in, boolean copy) {
return Nd4j.getExecutioner().exec(new Sinh(in, (copy ? in.ulike() : in)));
}
/**
* @param in
* @return
*/
public static INDArray cos(INDArray in) {
return cos(in, true);
}
/**
* @param in
* @param copy
* @return
*/
public static INDArray cosh(INDArray in, boolean copy) {
return Nd4j.getExecutioner().exec(new Cosh(in, (copy ? in.ulike() : in)));
}
/**
* @param in
* @return
*/
public static INDArray cosh(INDArray in) {
return cosh(in, true);
}
/**
* @param in
* @param copy
* @return
*/
public static INDArray cos(INDArray in, boolean copy) {
return Nd4j.getExecutioner().exec(new Cos(in, (copy ? in.ulike() : in)));
}
public static INDArray acos(INDArray arr) {
return acos(arr, true);
}
public static INDArray acos(INDArray in, boolean copy) {
return Nd4j.getExecutioner().exec(new ACos(in, (copy ? in.ulike() : in)));
}
public static INDArray asin(INDArray arr) {
return asin(arr, true);
}
public static INDArray asin(INDArray in, boolean copy) {
return Nd4j.getExecutioner().exec(new ASin(in, (copy ? in.ulike() : in)));
}
public static INDArray atan(INDArray arr) {
return atan(arr, true);
}
public static INDArray atan(INDArray in, boolean copy) {
return Nd4j.getExecutioner().exec(new ATan(in, (copy ? in.ulike() : in)));
}
public static INDArray ceil(INDArray arr) {
return ceil(arr, true);
}
public static INDArray ceil(INDArray in, boolean copy) {
return Nd4j.getExecutioner().exec(new Ceil(in, (copy ? in.ulike() : in)));
}
public static INDArray relu(INDArray arr) {
return relu(arr, true);
}
public static INDArray relu(INDArray in, boolean copy) {
return Nd4j.getExecutioner().exec(new RectifiedLinear(in, (copy ? in.ulike() : in)));
}
public static INDArray relu6(INDArray arr) {
return relu6(arr, true);
}
public static INDArray relu6(INDArray in, boolean copy) {
return Nd4j.getExecutioner().exec(new Relu6(in, (copy ? in.ulike() : in)));
}
public static INDArray leakyRelu(INDArray arr) {
return leakyRelu(arr, true);
}
public static INDArray leakyRelu(INDArray in, boolean copy) {
return Nd4j.getExecutioner().exec(new LeakyReLU(in, (copy ? in.ulike() : in)));
}
public static INDArray elu(INDArray arr) {
return elu(arr, true);
}
public static INDArray elu(INDArray in, boolean copy) {
return Nd4j.getExecutioner().exec(new ELU(in, (copy ? in.ulike() : in)))[0];
}
public static INDArray eluDerivative(INDArray arr, INDArray grad) {
return eluDerivative(arr, grad,true);
}
public static INDArray eluDerivative(INDArray in, INDArray grad, boolean copy) {
return Nd4j.getExecutioner().exec(new EluBp(in, grad, (copy ? in.ulike() : in)))[0];
}
public static INDArray leakyRelu(INDArray arr, double cutoff) {
return leakyRelu(arr, cutoff, true);
}
public static INDArray leakyRelu(INDArray in, double cutoff, boolean copy) {
return Nd4j.getExecutioner().exec(new LeakyReLU(in, (copy ? in.ulike() : in), cutoff));
}
public static INDArray leakyReluDerivative(INDArray arr, double cutoff) {
return leakyReluDerivative(arr, cutoff, true);
}
public static INDArray leakyReluDerivative(INDArray in, double cutoff, boolean copy) {
return Nd4j.getExecutioner().exec(new LeakyReLUDerivative(in, (copy ? in.ulike() : in), cutoff));
}
public static INDArray softPlus(INDArray arr) {
return softPlus(arr, true);
}
public static INDArray softPlus(INDArray in, boolean copy) {
return Nd4j.getExecutioner().exec(new SoftPlus(in, (copy ? in.ulike() : in)));
}
public static INDArray step(INDArray arr) {
return step(arr, true);
}
public static INDArray step(INDArray in, boolean copy) {
return Nd4j.getExecutioner().exec(new Step(in, (copy ? in.ulike() : in)));
}
public static INDArray softsign(INDArray arr) {
return softsign(arr, true);
}
public static INDArray softsign(INDArray in, boolean copy) {
return Nd4j.getExecutioner().exec(new SoftSign(in, (copy ? in.ulike() : in)));
}
public static INDArray softsignDerivative(INDArray arr) {
return softsignDerivative(arr, true);
}
public static INDArray softsignDerivative(INDArray in, boolean copy) {
return Nd4j.getExecutioner().exec(new SoftSignDerivative(in, (copy ? in.ulike() : in)));
}
public static INDArray softmax(INDArray arr) {
return softmax(arr, true);
}
/**
* @param in
* @param copy
* @return
*/
public static INDArray softmax(INDArray in, boolean copy) {
return Nd4j.getExecutioner().exec((CustomOp) new SoftMax(in, (copy ? in.ulike() : in), -1))[0];
}
/**
* out = in * (1-in)
*
* @param in Input array
* @param copy If true: copy. False: apply in-place
* @return
*/
public static INDArray timesOneMinus(INDArray in, boolean copy) {
return Nd4j.getExecutioner().exec(new TimesOneMinus(in, (copy ? in.ulike() : in)));
}
/**
* Abs function
*
* @param ndArray
* @return
*/
public static INDArray abs(INDArray ndArray) {
return abs(ndArray, true);
}
/**
* Run the exp operation
*
* @param ndArray
* @return
*/
public static INDArray exp(INDArray ndArray) {
return exp(ndArray, true);
}
public static INDArray hardTanh(INDArray ndArray) {
return hardTanh(ndArray, true);
}
/**
* Hard tanh
*
* @param ndArray the input
* @param dup whether to duplicate the ndarray and return it as the result
* @return the output
*/
public static INDArray hardTanh(INDArray ndArray, boolean dup) {
return exec(dup ? new HardTanh(ndArray, ndArray.ulike()) : new HardTanh(ndArray));
}
public static INDArray hardSigmoid(INDArray arr, boolean dup) {
return Nd4j.getExecutioner().exec(new HardSigmoid(arr, (dup ? arr.ulike() : arr)));
}
public static INDArray hardTanhDerivative(INDArray ndArray) {
return hardTanhDerivative(ndArray, true);
}
/**
* Hard tanh
*
* @param ndArray the input
* @param dup whether to duplicate the ndarray and return it as the result
* @return the output
*/
public static INDArray hardTanhDerivative(INDArray ndArray, boolean dup) {
return exec(dup ? new HardTanhDerivative(ndArray, ndArray.ulike()) : new HardTanhDerivative(ndArray));
}
/**
* @param ndArray
* @return
*/
public static INDArray identity(INDArray ndArray) {
return identity(ndArray, true);
}
/**
* Pow function
*
* @param ndArray the ndarray to raise hte power of
* @param power the power to raise by
* @return the ndarray raised to this power
*/
public static INDArray pow(INDArray ndArray, Number power) {
return pow(ndArray, power, true);
}
/**
* Element-wise power function - x^y, performed element-wise.
* Not performed in-place: the input arrays are not modified.
*
* @param ndArray the ndarray to raise to the power of
* @param power the power to raise by
* @return a copy of the ndarray raised to the specified power (element-wise)
*/
public static INDArray pow(INDArray ndArray, INDArray power) {
return pow(ndArray, power, true);
}
/**
* Element-wise power function - x^y, performed element-wise
*
* @param ndArray the ndarray to raise to the power of
* @param power the power to raise by
* @param dup if true:
* @return the ndarray raised to this power
*/
public static INDArray pow(INDArray ndArray, INDArray power, boolean dup) {
INDArray result = (dup ? ndArray.ulike() : ndArray);
return exec(new PowPairwise(ndArray, power, result));
}
/**
* Rounding function
*
* @param ndArray
* @return
*/
public static INDArray round(INDArray ndArray) {
return round(ndArray, true);
}
/**
* Sigmoid function
*
* @param ndArray
* @return
*/
public static INDArray sigmoid(INDArray ndArray) {
return sigmoid(ndArray, true);
}
/**
* Sigmoid function
*
* @param ndArray
* @param dup
* @return
*/
public static INDArray sigmoid(INDArray ndArray, boolean dup) {
return exec(dup ? new Sigmoid(ndArray, ndArray.ulike()) : new Sigmoid(ndArray));
}
/**
* Sigmoid function
*
* @param ndArray
* @return
*/
public static INDArray sigmoidDerivative(INDArray ndArray) {
return sigmoidDerivative(ndArray, true);
}
/**
* Sigmoid function
*
* @param ndArray
* @param dup
* @return
*/
public static INDArray sigmoidDerivative(INDArray ndArray, boolean dup) {
return exec(dup ? new SigmoidDerivative(ndArray, ndArray.ulike()) : new SigmoidDerivative(ndArray));
}
/**
* Sqrt function
*
* @param ndArray
* @return
*/
public static INDArray sqrt(INDArray ndArray) {
return sqrt(ndArray, true);
}
/**
* Element-wise tan function. Copies the array
*
* @param ndArray Input array
*/
public static INDArray tan(INDArray ndArray) {
return tan(ndArray, true);
}
/**
* Element-wise tan function. Copies the array
*
* @param ndArray Input array
*/
public static INDArray tan(INDArray ndArray, boolean dup) {
return exec(dup ? new Tan(ndArray, ndArray.ulike()) : new Tan(ndArray));
}
/**
* Tanh function
*
* @param ndArray
* @return
*/
public static INDArray tanh(INDArray ndArray) {
return tanh(ndArray, true);
}
/**
* Log on arbitrary base
*
* @param ndArray
* @param base
* @return
*/
public static INDArray log(INDArray ndArray, double base) {
return log(ndArray, base, true);
}
/**
* Log on arbitrary base
*
* @param ndArray
* @param base
* @return
*/
public static INDArray log(INDArray ndArray, double base, boolean duplicate) {
return Nd4j.getExecutioner().exec(new LogX(ndArray, duplicate ? ndArray.ulike() : ndArray, base));
}
public static INDArray log(INDArray ndArray) {
return log(ndArray, true);
}
public static INDArray eps(INDArray ndArray) {
return exec(new Eps(ndArray));
}
/**
* 1 if greater than or equal to 0 otherwise (at each element)
*
* @param first
* @param ndArray
* @return
*/
public static INDArray greaterThanOrEqual(INDArray first, INDArray ndArray) {
return greaterThanOrEqual(first, ndArray, true);
}
/**
* 1 if less than or equal to 0 otherwise (at each element)
*
* @param first
* @param ndArray
* @return
*/
public static INDArray lessThanOrEqual(INDArray first, INDArray ndArray) {
return lessThanOrEqual(first, ndArray, true);
}
/**
* Eps function
*
* @param ndArray
* @return
*/
public static INDArray lessThanOrEqual(INDArray first, INDArray ndArray, boolean dup) {
return Nd4j.getExecutioner().exec(new LessThanOrEqual(first, ndArray, Nd4j.createUninitialized(DataType.BOOL, first.shape(), first.ordering())))[0];
}
/**
* Eps function
*
* @param ndArray
* @return
*/
public static INDArray greaterThanOrEqual(INDArray first, INDArray ndArray, boolean dup) {
return Nd4j.getExecutioner().exec(new GreaterThanOrEqual(first, ndArray, Nd4j.createUninitialized(DataType.BOOL, first.shape(), first.ordering())))[0];
}
/**
* Floor function
*
* @param ndArray
* @return
*/
public static INDArray floor(INDArray ndArray, boolean dup) {
return exec(dup ? new Floor(ndArray, ndArray.ulike()) : new Floor(ndArray));
}
/**
* Signum function of this ndarray
*
* @param toSign
* @return
*/
public static INDArray sign(INDArray toSign, boolean dup) {
return exec(dup ? new Sign(toSign, toSign.ulike()) : new Sign(toSign));
}
/**
* Maximum function with a scalar
*
* @param ndArray tbe ndarray
* @param k
* @param dup
* @return
*/
public static INDArray max(INDArray ndArray, double k, boolean dup) {
return exec(dup ? new ScalarMax(ndArray, null, ndArray.ulike(), k) : new ScalarMax(ndArray, k));
}
/**
* Maximum function with a scalar
*
* @param ndArray tbe ndarray
* @param k
* @return
*/
public static INDArray max(INDArray ndArray, double k) {
return max(ndArray, k, true);
}
/**
* Element wise maximum function between 2 INDArrays
*
* @param first
* @param second
* @param dup
* @return
*/
public static INDArray max(INDArray first, INDArray second, boolean dup) {
long[] outShape = broadcastResultShape(first, second); //Also validates
Preconditions.checkState(dup || Arrays.equals(outShape, first.shape()), "Cannot do inplace max operation when first input is not equal to result shape (%ndShape vs. result %s)",
first, outShape);
INDArray out = dup ? Nd4j.create(first.dataType(), outShape) : first;
return Nd4j.exec(new org.nd4j.linalg.api.ops.impl.transforms.custom.Max(first, second, out))[0];
}
/**
* Element wise maximum function between 2 INDArrays
*
* @param first
* @param second
* @return
*/
public static INDArray max(INDArray first, INDArray second) {
return max(first, second, true);
}
/**
* Minimum function with a scalar
*
* @param ndArray tbe ndarray
* @param k
* @param dup
* @return
*/
public static INDArray min(INDArray ndArray, double k, boolean dup) {
return exec(dup ? new ScalarMin(ndArray, null, ndArray.ulike(), k) : new ScalarMin(ndArray, k));
}
/**
* Maximum function with a scalar
*
* @param ndArray tbe ndarray
* @param k
* @return
*/
public static INDArray min(INDArray ndArray, double k) {
return min(ndArray, k, true);
}
/**
* Element wise minimum function between 2 INDArrays
*
* @param first
* @param second
* @param dup
* @return
*/
public static INDArray min(INDArray first, INDArray second, boolean dup) {
long[] outShape = broadcastResultShape(first, second); //Also validates
Preconditions.checkState(dup || Arrays.equals(outShape, first.shape()), "Cannot do inplace min operation when first input is not equal to result shape (%ndShape vs. result %s)",
first, outShape);
INDArray out = dup ? Nd4j.create(first.dataType(), outShape) : first;
return Nd4j.exec(new org.nd4j.linalg.api.ops.impl.transforms.custom.Min(first, second, out))[0];
}
/**
* Element wise minimum function between 2 INDArrays
*
* @param first
* @param second
* @return
*/
public static INDArray min(INDArray first, INDArray second) {
return min(first, second, true);
}
/**
* Stabilize to be within a range of k
*
* @param ndArray tbe ndarray
* @param k
* @param dup
* @return
*/
public static INDArray stabilize(INDArray ndArray, double k, boolean dup) {
return exec(dup ? new Stabilize(ndArray, ndArray.ulike(), k) : new Stabilize(ndArray, k));
}
/**
* Abs function
*
* @param ndArray
* @param dup
* @return
*/
public static INDArray abs(INDArray ndArray, boolean dup) {
return exec(dup ? new Abs(ndArray, ndArray.ulike()) : new Abs(ndArray));
}
/**
* Exp function
*
* @param ndArray
* @param dup
* @return
*/
public static INDArray exp(INDArray ndArray, boolean dup) {
return exec(dup ? new Exp(ndArray, ndArray.ulike()) : new Exp(ndArray));
}
/**
* Elementwise exponential - 1 function
*
* @param ndArray
* @param dup
* @return
*/
public static INDArray expm1(INDArray ndArray, boolean dup) {
return exec(dup ? new Expm1(ndArray, ndArray.ulike()) : new Expm1(ndArray));
}
/**
* Identity function
*
* @param ndArray
* @param dup
* @return
*/
public static INDArray identity(INDArray ndArray, boolean dup) {
return Nd4j.getExecutioner().exec(dup ? new Identity(ndArray, ndArray.ulike()) : new Identity(ndArray, ndArray))[0];
}
public static INDArray isMax(INDArray input, DataType dataType) {
return isMax(input, Nd4j.createUninitialized(dataType, input.shape(), input.ordering()));
}
public static INDArray isMax(INDArray input) {
return isMax(input, input);
}
public static INDArray isMax(INDArray input, INDArray output) {
Nd4j.getExecutioner().exec(new IsMax(input, output));
return output;
}
/**
* Pow function
*
* @param ndArray
* @param power
* @param dup
* @return
*/
public static INDArray pow(INDArray ndArray, Number power, boolean dup) {
return exec(dup ? new Pow(ndArray, ndArray.ulike(), power.doubleValue()) : new Pow(ndArray, power.doubleValue()));
}
/**
* Rounding function
*
* @param ndArray the ndarray
* @param dup
* @return
*/
public static INDArray round(INDArray ndArray, boolean dup) {
return exec(dup ? new Round(ndArray, ndArray.ulike()) : new Round(ndArray));
}
/**
* Sqrt function
*
* @param ndArray
* @param dup
* @return
*/
public static INDArray sqrt(INDArray ndArray, boolean dup) {
return exec(dup ? new Sqrt(ndArray, ndArray.ulike()) : new Sqrt(ndArray, ndArray));
}
/**
* Tanh function
*
* @param ndArray
* @param dup
* @return
*/
public static INDArray tanh(INDArray ndArray, boolean dup) {
return exec(dup ? new Tanh(ndArray, ndArray.ulike()) : new Tanh(ndArray));
}
/**
* Log function
*
* @param ndArray
* @param dup
* @return
*/
public static INDArray log(INDArray ndArray, boolean dup) {
return exec(dup ? new Log(ndArray, ndArray.ulike()) : new Log(ndArray));
}
/**
* Log of x + 1 function
*
* @param ndArray
* @param dup
* @return
*/
public static INDArray log1p(INDArray ndArray, boolean dup) {
return exec(dup ? new Log1p(ndArray, ndArray.ulike()) : new Log1p(ndArray));
}
/**
* Negative
*
* @param ndArray
* @param dup
* @return
*/
public static INDArray neg(INDArray ndArray, boolean dup) {
return exec(dup ? new Negative(ndArray, ndArray.ulike()) : new Negative(ndArray));
}
public static INDArray and(INDArray x, INDArray y) {
INDArray z = Nd4j.createUninitialized(DataType.BOOL, x.shape(), x.ordering());
Nd4j.getExecutioner().exec(new And(x, y, z, 0.0));
return z;
}
public static INDArray or(INDArray x, INDArray y) {
INDArray z = Nd4j.createUninitialized(DataType.BOOL, x.shape(), x.ordering());
Nd4j.getExecutioner().exec(new Or(x, y, z, 0.0));
return z;
}
public static INDArray xor(INDArray x, INDArray y) {
INDArray z = Nd4j.createUninitialized(DataType.BOOL, x.shape(), x.ordering());
Nd4j.getExecutioner().exec(new Xor(x, y, z, 0.0));
return z;
}
public static INDArray not(INDArray x) {
val z = Nd4j.createUninitialized(DataType.BOOL, x.shape(), x.ordering());
if (x.isB()) {
Nd4j.getExecutioner().exec(new BooleanNot(x, z));
} else {
Nd4j.getExecutioner().exec(new ScalarNot(x, z, 0.0f));
}
return z;
}
/**
* Apply the given elementwise op
*
* @param op the factory to create the op
* @return the new ndarray
*/
private static INDArray exec(ScalarOp op) {
return Nd4j.getExecutioner().exec(op);
}
/**
* Apply the given elementwise op
*
* @param op the factory to create the op
* @return the new ndarray
*/
private static INDArray exec(TransformOp op) {
return Nd4j.getExecutioner().exec(op);
}
/**
* Raises a square matrix to a power n, which can be positive, negative, or zero.
* The behavior is similar to the numpy matrix_power() function. The algorithm uses
* repeated squarings to minimize the number of mmul() operations needed
* If n is zero, the identity matrix is returned.
* If n is negative, the matrix is inverted and raised to the abs(n) power.
*
* @param in A square matrix to raise to an integer power, which will be changed if dup is false.
* @param n The integer power to raise the matrix to.
* @param dup If dup is true, the original input is unchanged.
* @return The result of raising in to the nth power.
*/
public static INDArray mpow(INDArray in, int n, boolean dup) {
Preconditions.checkState(in.isMatrix() && in.isSquare(), "Input must be a square matrix: got input with shape %s", in.shape());
if (n == 0) {
if (dup)
return Nd4j.eye(in.rows());
else
return in.assign(Nd4j.eye(in.rows()));
}
INDArray temp;
if (n < 0) {
temp = InvertMatrix.invert(in, !dup);
n = -n;
} else
temp = in.dup();
INDArray result = temp.dup();
if (n < 4) {
for (int i = 1; i < n; i++) {
result.mmuli(temp);
}
if (dup)
return result;
else
return in.assign(result);
} else {
// lets try to optimize by squaring itself a bunch of times
int squares = (int) (Math.log(n) / Math.log(2.0));
for (int i = 0; i < squares; i++)
result = result.mmul(result);
int diff = (int) Math.round(n - Math.pow(2.0, squares));
for (int i = 0; i < diff; i++)
result.mmuli(temp);
if (dup)
return result;
else
return in.assign(result);
}
}
protected static long[] broadcastResultShape(INDArray first, INDArray second){
if(first.equalShapes(second)){
return first.shape();
} else if(Shape.areShapesBroadcastable(first.shape(), second.shape())){
return Shape.broadcastOutputShape(first.shape(), second.shape());
} else {
throw new IllegalStateException("Array shapes are not broadcastable: " + Arrays.toString(first.shape()) +
" vs. " + Arrays.toString(second.shape()));
}
}
}