Package org.nd4j.linalg.factory.ops
Class NDMath
- java.lang.Object
-
- org.nd4j.linalg.factory.ops.NDMath
-
public class NDMath extends Object
-
-
Constructor Summary
Constructors Constructor Description NDMath()
-
Method Summary
All Methods Instance Methods Concrete Methods Modifier and Type Method Description INDArray
abs(INDArray x)
Elementwise absolute value operation: out = abs(x)INDArray
acos(INDArray x)
Elementwise acos (arccosine, inverse cosine) operation: out = arccos(x)INDArray
acosh(INDArray x)
Elementwise acosh (inverse hyperbolic cosine) function: out = acosh(x)INDArray
add(INDArray x, double value)
Scalar add operation, out = in + scalarINDArray
add(INDArray x, INDArray y)
Pairwise addition operation, out = x + y
Note: supports broadcasting if x and y have different shapes and are broadcastable.
For example, if X has shape [1,10] and Y has shape [5,10] then op(X,Y) has output shape [5,10]
Broadcast rules are the same as NumPy: https://docs.scipy.org/doc/numpy/user/basics.broadcasting.htmlINDArray
and(INDArray x, INDArray y)
Boolean AND operation: elementwise (x != 0) && (y != 0)
If x and y arrays have equal shape, the output shape is the same as these inputs.
Note: supports broadcasting if x and y have different shapes and are broadcastable.
Returns an array with values 1 where condition is satisfied, or value 0 otherwise.INDArray
asin(INDArray x)
Elementwise asin (arcsin, inverse sine) operation: out = arcsin(x)INDArray
asinh(INDArray x)
Elementwise asinh (inverse hyperbolic sine) function: out = asinh(x)INDArray
asum(INDArray in, boolean keepDims, int... dimensions)
Absolute sum array reduction operation, optionally along specified dimensions: out = sum(abs(x))INDArray
asum(INDArray in, int... dimensions)
Absolute sum array reduction operation, optionally along specified dimensions: out = sum(abs(x))INDArray
atan(INDArray x)
Elementwise atan (arctangent, inverse tangent) operation: out = arctangent(x)INDArray
atan2(INDArray y, INDArray x)
Elementwise atan (arctangent, inverse tangent) operation: out = atan2(x,y).
Similar to atan(y/x) but sigts of x and y are used to determine the location of the resultINDArray
atanh(INDArray x)
Elementwise atanh (inverse hyperbolic tangent) function: out = atanh(x)INDArray
bitShift(INDArray x, INDArray shift)
Bit shift operationINDArray
bitShiftRight(INDArray x, INDArray shift)
Right bit shift operationINDArray
bitShiftRotl(INDArray x, INDArray shift)
Cyclic bit shift operationINDArray
bitShiftRotr(INDArray x, INDArray shift)
Cyclic right shift operationINDArray
ceil(INDArray x)
Element-wise ceiling function: out = ceil(x).
Rounds each value up to the nearest integer value (if not already an integer)INDArray
clipByAvgNorm(INDArray x, double clipValue, int... dimensions)
Clips tensor values to a maximum average L2-norm.INDArray
clipByNorm(INDArray x, double clipValue, int... dimensions)
Clipping by L2 norm, optionally along dimension(s)
if l2Norm(x,dimension) < clipValue, then input is returned unmodifed
Otherwise, out[i] = in[i] * clipValue / l2Norm(in, dimensions) where each value is clipped according
to the corresponding l2Norm along the specified dimensionsINDArray
clipByValue(INDArray x, double clipValueMin, double clipValueMax)
Element-wise clipping function:
out[i] = in[i] if in[i] >= clipValueMin and in[i] <= clipValueMax
out[i] = clipValueMin if in[i] < clipValueMin
out[i] = clipValueMax if in[i] > clipValueMaxINDArray
confusionMatrix(INDArray labels, INDArray pred, int numClasses)
Compute the 2d confusion matrix of size [numClasses, numClasses] from a pair of labels and predictions, both of
which are represented as integer values.
For example, if labels = [0, 1, 1], predicted = [0, 2, 1], and numClasses=4 then output is:
[1, 0, 0, 0]
[0, 1, 1, 0]
[0, 0, 0, 0]
[0, 0, 0, 0]INDArray
confusionMatrix(INDArray labels, INDArray pred, DataType dataType)
Compute the 2d confusion matrix of size [numClasses, numClasses] from a pair of labels and predictions, both of
which are represented as integer values.INDArray
confusionMatrix(INDArray labels, INDArray pred, INDArray weights)
Compute the 2d confusion matrix of size [numClasses, numClasses] from a pair of labels and predictions, both of
which are represented as integer values.INDArray
confusionMatrix(INDArray labels, INDArray pred, INDArray weights, int numClasses)
Compute the 2d confusion matrix of size [numClasses, numClasses] from a pair of labels and predictions, both of
which are represented as integer values.
For example, if labels = [0, 1, 1], predicted = [0, 2, 1], numClasses = 4, and weights = [1, 2, 3]
[1, 0, 0, 0]
[0, 3, 2, 0]
[0, 0, 0, 0]
[0, 0, 0, 0]INDArray
cos(INDArray x)
Elementwise cosine operation: out = cos(x)INDArray
cosh(INDArray x)
Elementwise cosh (hyperbolic cosine) operation: out = cosh(x)INDArray
cosineDistance(INDArray x, INDArray y, boolean keepDims, boolean isComplex, int... dimensions)
Cosine distance reduction operation.INDArray
cosineDistance(INDArray x, INDArray y, int... dimensions)
Cosine distance reduction operation.INDArray
cosineSimilarity(INDArray x, INDArray y, boolean keepDims, boolean isComplex, int... dimensions)
Cosine similarity pairwise reduction operation.INDArray
cosineSimilarity(INDArray x, INDArray y, int... dimensions)
Cosine similarity pairwise reduction operation.INDArray
countNonZero(INDArray in, boolean keepDims, int... dimensions)
Count non zero array reduction operation, optionally along specified dimensions: out = count(x != 0)INDArray
countNonZero(INDArray in, int... dimensions)
Count non zero array reduction operation, optionally along specified dimensions: out = count(x != 0)INDArray
countZero(INDArray in, boolean keepDims, int... dimensions)
Count zero array reduction operation, optionally along specified dimensions: out = count(x == 0)INDArray
countZero(INDArray in, int... dimensions)
Count zero array reduction operation, optionally along specified dimensions: out = count(x == 0)INDArray
cross(INDArray a, INDArray b)
Returns the pair-wise cross product of equal size arrays a and b: a x b = ||a||x||b|| sin(theta).
Can take rank 1 or above inputs (of equal shapes), but note that the last dimension must have dimension 3INDArray
cube(INDArray x)
Element-wise cube function: out = x^3INDArray
diag(INDArray x)
Returns an output variable with diagonal values equal to the specified values; off-diagonal values will be set to 0
For example, if input = [1,2,3], then output is given by:
[ 1, 0, 0]
[ 0, 2, 0]
[ 0, 0, 3]
Higher input ranks are also supported: if input has shape [a,...,R-1] then output[i,...,k,i,...,k] = input[i,...,k].
i.e., for input rank R, output has rank 2RINDArray
diagPart(INDArray x)
Extract the diagonal part from the input array.
If input is
[ 1, 0, 0]
[ 0, 2, 0]
[ 0, 0, 3]
then output is [1, 2, 3].
Supports higher dimensions: in general, out[i,...,k] = in[i,...,k,i,...,k]INDArray
div(INDArray x, double value)
Scalar division operation, out = in / scalarINDArray
div(INDArray x, INDArray y)
Pairwise division operation, out = x / y
Note: supports broadcasting if x and y have different shapes and are broadcastable.
For example, if X has shape [1,10] and Y has shape [5,10] then op(X,Y) has output shape [5,10]
Broadcast rules are the same as NumPy: https://docs.scipy.org/doc/numpy/user/basics.broadcasting.htmlINDArray
embeddingLookup(INDArray x, INDArray indices, PartitionMode PartitionMode)
Looks up ids in a list of embedding tensors.INDArray
entropy(INDArray in, boolean keepDims, int... dimensions)
Entropy reduction: -sum(x * log(x))INDArray
entropy(INDArray in, int... dimensions)
Entropy reduction: -sum(x * log(x))INDArray
erf(INDArray x)
Element-wise Gaussian error function - out = erf(in)INDArray
erfc(INDArray x)
Element-wise complementary Gaussian error function - out = erfc(in) = 1 - erf(in)INDArray
euclideanDistance(INDArray x, INDArray y, boolean keepDims, boolean isComplex, int... dimensions)
Euclidean distance (l2 norm, l2 distance) reduction operation.INDArray
euclideanDistance(INDArray x, INDArray y, int... dimensions)
Euclidean distance (l2 norm, l2 distance) reduction operation.INDArray
exp(INDArray x)
Elementwise exponent function: out = exp(x) = 2.71828...^xINDArray
expm1(INDArray x)
Elementwise 1.0 - exponent function: out = 1.0 - exp(x) = 1.0 - 2.71828...^xINDArray
eye(int rows)
Generate an identity matrix with the specified number of rows and columns.INDArray
eye(int rows, int cols)
As per eye(String, int, int, DataType) but with the default datatype, Eye.DEFAULT_DTYPEINDArray
eye(int rows, int cols, DataType dataType, int... dimensions)
Generate an identity matrix with the specified number of rows and columns
Example:INDArray
eye(INDArray rows)
As per eye(String, int) but with the number of rows specified as a scalar INDArrayINDArray
eye(INDArray rows, INDArray cols)
As per eye(int, int) bit with the number of rows/columns specified as scalar INDArraysINDArray
firstIndex(INDArray in, Condition condition, boolean keepDims, int... dimensions)
First index reduction operation.
Returns a variable that contains the index of the first element that matches the specified condition (for each
slice along the specified dimensions)
Note that if keepDims = true, the output variable has the same rank as the input variable,
with the reduced dimensions having size 1.INDArray
firstIndex(INDArray in, Condition condition, int... dimensions)
First index reduction operation.
Returns a variable that contains the index of the first element that matches the specified condition (for each
slice along the specified dimensions)
Note that if keepDims = true, the output variable has the same rank as the input variable,
with the reduced dimensions having size 1.INDArray
floor(INDArray x)
Element-wise floor function: out = floor(x).
Rounds each value down to the nearest integer value (if not already an integer)INDArray
floorDiv(INDArray x, INDArray y)
Pairwise floor division operation, out = floor(x / y)
Note: supports broadcasting if x and y have different shapes and are broadcastable.
For example, if X has shape [1,10] and Y has shape [5,10] then op(X,Y) has output shape [5,10]
Broadcast rules are the same as NumPy: https://docs.scipy.org/doc/numpy/user/basics.broadcasting.htmlINDArray
floorMod(INDArray x, double value)
Scalar floor modulus operationINDArray
floorMod(INDArray x, INDArray y)
Pairwise Modulus division operation
Note: supports broadcasting if x and y have different shapes and are broadcastable.
For example, if X has shape [1,10] and Y has shape [5,10] then op(X,Y) has output shape [5,10]
Broadcast rules are the same as NumPy: https://docs.scipy.org/doc/numpy/user/basics.broadcasting.htmlINDArray
hammingDistance(INDArray x, INDArray y, boolean keepDims, boolean isComplex, int... dimensions)
Hamming distance reduction operation.INDArray
hammingDistance(INDArray x, INDArray y, int... dimensions)
Hamming distance reduction operation.INDArray
iamax(INDArray in, boolean keepDims, int... dimensions)
Index of the max absolute value: argmax(abs(in))
see argmax(String, INDArray, boolean, int...)INDArray
iamax(INDArray in, int... dimensions)
Index of the max absolute value: argmax(abs(in))
see argmax(String, INDArray, boolean, int...)INDArray
iamin(INDArray in, boolean keepDims, int... dimensions)
Index of the min absolute value: argmin(abs(in))
see argmin(String, INDArray, boolean, int...)INDArray
iamin(INDArray in, int... dimensions)
Index of the min absolute value: argmin(abs(in))
see argmin(String, INDArray, boolean, int...)INDArray
isFinite(INDArray x)
Is finite operation: elementwise isFinite(x)
Returns an array with the same shape/size as the input, with values 1 where condition is satisfied, or
value 0 otherwiseINDArray
isInfinite(INDArray x)
Is infinite operation: elementwise isInfinite(x)
Returns an array with the same shape/size as the input, with values 1 where condition is satisfied, or
value 0 otherwiseINDArray
isMax(INDArray x)
Is maximum operation: elementwise x == max(x)
Returns an array with the same shape/size as the input, with values 1 where condition is satisfied, or
value 0 otherwiseINDArray
isNaN(INDArray x)
Is Not a Number operation: elementwise isNaN(x)
Returns an array with the same shape/size as the input, with values 1 where condition is satisfied, or
value 0 otherwiseINDArray
isNonDecreasing(INDArray x)
Is the array non decreasing?
An array is non-decreasing if for every valid i, x[i] <= x[i+1].INDArray
isStrictlyIncreasing(INDArray x)
Is the array strictly increasing?
An array is strictly increasing if for every valid i, x[i] < x[i+1].INDArray
jaccardDistance(INDArray x, INDArray y, boolean keepDims, boolean isComplex, int... dimensions)
Jaccard similarity reduction operation.INDArray
jaccardDistance(INDArray x, INDArray y, int... dimensions)
Jaccard similarity reduction operation.INDArray
lastIndex(INDArray in, Condition condition, boolean keepDims, int... dimensions)
Last index reduction operation.
Returns a variable that contains the index of the last element that matches the specified condition (for each
slice along the specified dimensions)
Note that if keepDims = true, the output variable has the same rank as the input variable,
with the reduced dimensions having size 1.INDArray
lastIndex(INDArray in, Condition condition, int... dimensions)
Last index reduction operation.
Returns a variable that contains the index of the last element that matches the specified condition (for each
slice along the specified dimensions)
Note that if keepDims = true, the output variable has the same rank as the input variable,
with the reduced dimensions having size 1.INDArray[]
listDiff(INDArray x, INDArray y)
Calculates difference between inputs X and Y.INDArray
log(INDArray x)
Element-wise logarithm function (base e - natural logarithm): out = log(x)INDArray
log(INDArray x, double base)
Element-wise logarithm function (with specified base): out = log_{base}(x)INDArray
log1p(INDArray x)
Elementwise natural logarithm function: out = log_e (1 + x)INDArray
logEntropy(INDArray in, boolean keepDims, int... dimensions)
Log entropy reduction: log(-sum(x * log(x)))INDArray
logEntropy(INDArray in, int... dimensions)
Log entropy reduction: log(-sum(x * log(x)))INDArray
logSumExp(INDArray input, int... dimensions)
Log-sum-exp reduction (optionally along dimension).
Computes log(sum(exp(x))INDArray
manhattanDistance(INDArray x, INDArray y, boolean keepDims, boolean isComplex, int... dimensions)
Manhattan distance (l1 norm, l1 distance) reduction operation.INDArray
manhattanDistance(INDArray x, INDArray y, int... dimensions)
Manhattan distance (l1 norm, l1 distance) reduction operation.INDArray
matrixDeterminant(INDArray in)
Matrix determinant op.INDArray
matrixInverse(INDArray in)
Matrix inverse op.INDArray
max(INDArray x, INDArray y)
Pairwise max operation, out = max(x, y)
Note: supports broadcasting if x and y have different shapes and are broadcastable.
For example, if X has shape [1,10] and Y has shape [5,10] then op(X,Y) has output shape [5,10]
Broadcast rules are the same as NumPy: https://docs.scipy.org/doc/numpy/user/basics.broadcasting.htmlINDArray
mean(INDArray in, boolean keepDims, int... dimensions)
Mean array reduction operation, optionally along specified dimensions: out = mean(abs(x))INDArray
mean(INDArray in, int... dimensions)
Mean array reduction operation, optionally along specified dimensions: out = mean(abs(x))INDArray
mean(INDArray in, INDArray dimensions)
Mean array reduction operation, optionally along specified dimensions: out = mean(abs(x))INDArray
mean(INDArray in, INDArray dimensions, boolean keepDims)
Mean array reduction operation, optionally along specified dimensions: out = mean(abs(x))INDArray
mergeAdd(INDArray... inputs)
Merge add function: merges an arbitrary number of equal shaped arrays using element-wise addition:
out = sum_i in[i]INDArray
mergeAvg(INDArray... inputs)
Merge average function: merges an arbitrary number of equal shaped arrays using element-wise mean operation:
out = mean_i in[i]INDArray
mergeMax(INDArray... inputs)
Merge max function: merges an arbitrary number of equal shaped arrays using element-wise maximum operation:
out = max_i in[i]INDArray
mergeMaxIndex(INDArray... x)
Return array of max elements indices with along tensor dimensionsINDArray
mergeMaxIndex(INDArray[] x, DataType dataType)
Return array of max elements indices with along tensor dimensionsINDArray[]
meshgrid(INDArray[] inputs, boolean cartesian)
Broadcasts parameters for evaluation on an N-D grid.INDArray
min(INDArray x, INDArray y)
Pairwise max operation, out = min(x, y)
Note: supports broadcasting if x and y have different shapes and are broadcastable.
For example, if X has shape [1,10] and Y has shape [5,10] then op(X,Y) has output shape [5,10]
Broadcast rules are the same as NumPy: https://docs.scipy.org/doc/numpy/user/basics.broadcasting.htmlINDArray
mod(INDArray x, INDArray y)
Pairwise modulus (remainder) operation, out = x % y
Note: supports broadcasting if x and y have different shapes and are broadcastable.
For example, if X has shape [1,10] and Y has shape [5,10] then op(X,Y) has output shape [5,10]
Broadcast rules are the same as NumPy: https://docs.scipy.org/doc/numpy/user/basics.broadcasting.htmlINDArray[]
moments(INDArray input, int[] axes, boolean keepDims)
Calculate the mean and (population) variance for the input variable, for the specified axisINDArray[]
moments(INDArray input, INDArray axes, boolean keepDims)
Calculate the mean and (population) variance for the input variable, for the specified axisINDArray
mul(INDArray x, double value)
Scalar multiplication operation, out = in * scalarINDArray
mul(INDArray x, INDArray y)
Pairwise multiplication operation, out = x * y
Note: supports broadcasting if x and y have different shapes and are broadcastable.
For example, if X has shape [1,10] and Y has shape [5,10] then op(X,Y) has output shape [5,10]
Broadcast rules are the same as NumPy: https://docs.scipy.org/doc/numpy/user/basics.broadcasting.htmlINDArray
neg(INDArray x)
Elementwise negative operation: out = -xINDArray
norm1(INDArray in, boolean keepDims, int... dimensions)
Mean array reduction operation, optionally along specified dimensions: out = mean(abs(x))INDArray
norm1(INDArray in, int... dimensions)
Mean array reduction operation, optionally along specified dimensions: out = mean(abs(x))INDArray
norm1(INDArray in, INDArray dimensions)
Sum of absolute differences.INDArray
norm1(INDArray in, INDArray dimensions, boolean keepDims)
Sum of absolute differences.INDArray
norm2(INDArray in, boolean keepDims, int... dimensions)
Euclidean norm: euclidean distance of a vector from the originINDArray
norm2(INDArray in, int... dimensions)
Euclidean norm: euclidean distance of a vector from the originINDArray
norm2(INDArray in, INDArray dimensions)
Euclidean norm: euclidean distance of a vector from the originINDArray
norm2(INDArray in, INDArray dimensions, boolean keepDims)
Euclidean norm: euclidean distance of a vector from the originINDArray[]
normalizeMoments(INDArray counts, INDArray means, INDArray variances, double shift)
Calculate the mean and variance from the sufficient statisticsINDArray
normMax(INDArray in, boolean keepDims, int... dimensions)
Differences between max absolute valueINDArray
normMax(INDArray in, int... dimensions)
Differences between max absolute valueINDArray
normMax(INDArray in, INDArray dimensions)
Differences between max absolute valueINDArray
normMax(INDArray in, INDArray dimensions, boolean keepDims)
Differences between max absolute valueINDArray
or(INDArray x, INDArray y)
Boolean OR operation: elementwise (x != 0) || (y != 0)
If x and y arrays have equal shape, the output shape is the same as these inputs.
Note: supports broadcasting if x and y have different shapes and are broadcastable.
Returns an array with values 1 where condition is satisfied, or value 0 otherwise.INDArray
pow(INDArray x, double value)
Element-wise power function: out = x^valueINDArray
pow(INDArray x, INDArray y)
Element-wise (broadcastable) power function: out = x[i]^y[i]INDArray
prod(INDArray in, boolean keepDims, int... dimensions)
The max of an array along each dimensionINDArray
prod(INDArray in, int... dimensions)
The max of an array along each dimensionINDArray
prod(INDArray in, INDArray dimensions)
The product of an array long each dimensionINDArray
prod(INDArray in, INDArray dimensions, boolean keepDims)
The product of an array long each dimensionINDArray
rationalTanh(INDArray x)
Rational Tanh Approximation elementwise function, as described in the paper:
Compact Convolutional Neural Network Cascade for Face Detection
This is a faster Tanh approximationINDArray
rdiv(INDArray x, double value)
Scalar reverse division operation, out = scalar / inINDArray
rdiv(INDArray x, INDArray y)
Pairwise reverse division operation, out = y / x
Note: supports broadcasting if x and y have different shapes and are broadcastable.
For example, if X has shape [1,10] and Y has shape [5,10] then op(X,Y) has output shape [5,10]
Broadcast rules are the same as NumPy: https://docs.scipy.org/doc/numpy/user/basics.broadcasting.htmlINDArray
reciprocal(INDArray x)
Element-wise reciprocal (inverse) function: out[i] = 1 / in[i]INDArray
rectifiedTanh(INDArray x)
Rectified tanh operation: max(0, tanh(in))INDArray
reduceAMax(INDArray in, boolean keepDims, int... dimensions)
Absolute max array reduction operation, optionally along specified dimensions: out = max(abs(x))INDArray
reduceAMax(INDArray in, int... dimensions)
Absolute max array reduction operation, optionally along specified dimensions: out = max(abs(x))INDArray
reduceAMax(INDArray in, INDArray dimensions)
Absolute max array reduction operation, optionally along specified dimensions: out = max(abs(x))INDArray
reduceAMax(INDArray in, INDArray dimensions, boolean keepDims)
Absolute max array reduction operation, optionally along specified dimensions: out = max(abs(x))INDArray
reduceAmean(INDArray in, boolean keepDims, int... dimensions)
Absolute mean array reduction operation, optionally along specified dimensions: out = mean(abs(x))INDArray
reduceAmean(INDArray in, int... dimensions)
Absolute mean array reduction operation, optionally along specified dimensions: out = mean(abs(x))INDArray
reduceAmean(INDArray in, INDArray dimensions)
Absolute mean array reduction operation, optionally along specified dimensions: out = mean(abs(x))INDArray
reduceAmean(INDArray in, INDArray dimensions, boolean keepDims)
Absolute mean array reduction operation, optionally along specified dimensions: out = mean(abs(x))INDArray
reduceAmin(INDArray in, boolean keepDims, int... dimensions)
Absolute min array reduction operation, optionally along specified dimensions: out = min(abs(x))INDArray
reduceAmin(INDArray in, int... dimensions)
Absolute min array reduction operation, optionally along specified dimensions: out = min(abs(x))INDArray
reduceAmin(INDArray in, INDArray dimensions)
Absolute min array reduction operation, optionally along specified dimensions: out = min(abs(x))INDArray
reduceAmin(INDArray in, INDArray dimensions, boolean keepDims)
Absolute min array reduction operation, optionally along specified dimensions: out = min(abs(x))INDArray
reduceMax(INDArray in, boolean keepDims, int... dimensions)
The max of an array along each dimensionINDArray
reduceMax(INDArray in, int... dimensions)
The max of an array along each dimensionINDArray
reduceMax(INDArray in, INDArray dimensions)
The max of an array long each dimensionINDArray
reduceMax(INDArray in, INDArray dimensions, boolean keepDims)
The max of an array long each dimensionINDArray
reduceMin(INDArray in, boolean keepDims, int... dimensions)
The minimum of an array along each dimensionINDArray
reduceMin(INDArray in, int... dimensions)
The minimum of an array along each dimensionINDArray
reduceMin(INDArray in, INDArray dimensions)
The minimum of an array long each dimensionINDArray
reduceMin(INDArray in, INDArray dimensions, boolean keepDims)
The minimum of an array long each dimensionINDArray
round(INDArray x)
Element-wise round function: out = round(x).
Rounds (up or down depending on value) to the nearest integer value.INDArray
rsqrt(INDArray x)
Element-wise reciprocal (inverse) of square root: out = 1.0 / sqrt(x)INDArray
rsub(INDArray x, double value)
Scalar reverse subtraction operation, out = scalar - inINDArray
rsub(INDArray x, INDArray y)
Pairwise reverse subtraction operation, out = y - x
Note: supports broadcasting if x and y have different shapes and are broadcastable.
For example, if X has shape [1,10] and Y has shape [5,10] then op(X,Y) has output shape [5,10]
Broadcast rules are the same as NumPy: https://docs.scipy.org/doc/numpy/user/basics.broadcasting.htmlINDArray
setDiag(INDArray in, INDArray diag)
Set the diagonal value to the specified values
If input is
[ a, b, c]
[ d, e, f]
[ g, h, i]
and diag = [ 1, 2, 3] then output is
[ 1, b, c]
[ d, 2, f]
[ g, h, 3]INDArray
shannonEntropy(INDArray in, boolean keepDims, int... dimensions)
Shannon Entropy reduction: -sum(x * log2(x))INDArray
shannonEntropy(INDArray in, int... dimensions)
Shannon Entropy reduction: -sum(x * log2(x))INDArray
shannonEntropy(INDArray in, INDArray dimensions)
Shannon Entropy reduction: -sum(x * log2(x))INDArray
shannonEntropy(INDArray in, INDArray dimensions, boolean keepDims)
Shannon Entropy reduction: -sum(x * log2(x))INDArray
sign(INDArray x)
Element-wise sign (signum) function:
out = -1 if in < 0
out = 0 if in = 0
out = 1 if in > 0INDArray
sin(INDArray x)
Elementwise sine operation: out = sin(x)INDArray
sinh(INDArray x)
Elementwise sinh (hyperbolic sine) operation: out = sinh(x)INDArray
sqrt(INDArray x)
Element-wise square root function: out = sqrt(x)INDArray
square(INDArray x)
Element-wise square function: out = x^2INDArray
squaredDifference(INDArray x, INDArray y)
Pairwise squared difference operation.
Note: supports broadcasting if x and y have different shapes and are broadcastable.
For example, if X has shape [1,10] and Y has shape [5,10] then op(X,Y) has output shape [5,10]
Broadcast rules are the same as NumPy: https://docs.scipy.org/doc/numpy/user/basics.broadcasting.htmlINDArray
squaredNorm(INDArray in, boolean keepDims, int... dimensions)
Sum of squared differences.INDArray
squaredNorm(INDArray in, int... dimensions)
Sum of squared differences.INDArray
squaredNorm(INDArray in, INDArray dimensions)
Sum of squared differences.INDArray
squaredNorm(INDArray in, INDArray dimensions, boolean keepDims)
Sum of squared differences.INDArray
standardize(INDArray x, int... dimensions)
Standardize input variable along given axisINDArray
step(INDArray x, double value)
Elementwise step function:
out(x) = 1 if x >= cutoff
out(x) = 0 otherwiseINDArray
sub(INDArray x, double value)
Scalar subtraction operation, out = in - scalarINDArray
sub(INDArray x, INDArray y)
Pairwise subtraction operation, out = x - y
Note: supports broadcasting if x and y have different shapes and are broadcastable.
For example, if X has shape [1,10] and Y has shape [5,10] then op(X,Y) has output shape [5,10]
Broadcast rules are the same as NumPy: https://docs.scipy.org/doc/numpy/user/basics.broadcasting.htmlINDArray
sum(INDArray in, boolean keepDims, int... dimensions)
Sum of an array, optionally along specified dimensions: out = sum(x))INDArray
sum(INDArray in, int... dimensions)
Sum of an array, optionally along specified dimensions: out = sum(x))INDArray
sum(INDArray in, INDArray dimensions)
Sum of an array, optionally along specified dimensions: out = sum(x))INDArray
sum(INDArray in, INDArray dimensions, boolean keepDims)
Sum of an array, optionally along specified dimensions: out = sum(x))INDArray
tan(INDArray x)
Elementwise tangent operation: out = tan(x)INDArray
tanh(INDArray x)
Elementwise tanh (hyperbolic tangent) operation: out = tanh(x)INDArray
trace(INDArray in)
Matrix trace operation
For rank 2 matrices, the output is a scalar with the trace - i.e., sum of the main diagonal.
For higher rank inputs, output[a,b,c] = trace(in[a,b,c,:,:])INDArray
xor(INDArray x, INDArray y)
Boolean XOR (exclusive OR) operation: elementwise (x != 0) XOR (y != 0)
If x and y arrays have equal shape, the output shape is the same as these inputs.
Note: supports broadcasting if x and y have different shapes and are broadcastable.
Returns an array with values 1 where condition is satisfied, or value 0 otherwise.INDArray
zeroFraction(INDArray input)
Full array zero fraction array reduction operation, optionally along specified dimensions: out = (count(x == 0) / length(x))
-
-
-
Method Detail
-
clipByAvgNorm
public INDArray clipByAvgNorm(INDArray x, double clipValue, int... dimensions)
Clips tensor values to a maximum average L2-norm.- Parameters:
x
- Input variable (NUMERIC type)clipValue
- Value for clippingdimensions
- Dimensions to reduce over (Size: AtLeast(min=0))- Returns:
- output Output variable (NUMERIC type)
-
embeddingLookup
public INDArray embeddingLookup(INDArray x, INDArray indices, PartitionMode PartitionMode)
Looks up ids in a list of embedding tensors.- Parameters:
x
- Input tensor (NUMERIC type)indices
- A Tensor containing the ids to be looked up. (INT type)PartitionMode
- partition_mode == 0 - i.e. 'mod' , 1 - 'div'- Returns:
- output Shifted output (NUMERIC type)
-
mergeMaxIndex
public INDArray mergeMaxIndex(INDArray[] x, DataType dataType)
Return array of max elements indices with along tensor dimensions- Parameters:
x
- Input tensor (NUMERIC type)dataType
- Data type- Returns:
- output Array max elements indices with along dimensions. (INT type)
-
mergeMaxIndex
public INDArray mergeMaxIndex(INDArray... x)
Return array of max elements indices with along tensor dimensions- Parameters:
x
- Input tensor (NUMERIC type)- Returns:
- output Array max elements indices with along dimensions. (INT type)
-
abs
public INDArray abs(INDArray x)
Elementwise absolute value operation: out = abs(x)- Parameters:
x
- Input variable (NUMERIC type)- Returns:
- output Output variable (NUMERIC type)
-
acos
public INDArray acos(INDArray x)
Elementwise acos (arccosine, inverse cosine) operation: out = arccos(x)- Parameters:
x
- Input variable (NUMERIC type)- Returns:
- output Output variable (NUMERIC type)
-
acosh
public INDArray acosh(INDArray x)
Elementwise acosh (inverse hyperbolic cosine) function: out = acosh(x)- Parameters:
x
- Input variable (NUMERIC type)- Returns:
- output Output variable (NUMERIC type)
-
add
public INDArray add(INDArray x, INDArray y)
Pairwise addition operation, out = x + y
Note: supports broadcasting if x and y have different shapes and are broadcastable.
For example, if X has shape [1,10] and Y has shape [5,10] then op(X,Y) has output shape [5,10]
Broadcast rules are the same as NumPy: https://docs.scipy.org/doc/numpy/user/basics.broadcasting.html- Parameters:
x
- Input variable (NUMERIC type)y
- Input variable (NUMERIC type)- Returns:
- output Output variable (NUMERIC type)
-
add
public INDArray add(INDArray x, double value)
Scalar add operation, out = in + scalar- Parameters:
x
- Input variable (NUMERIC type)value
- Scalar value for op- Returns:
- output Output variable (NUMERIC type)
-
and
public INDArray and(INDArray x, INDArray y)
Boolean AND operation: elementwise (x != 0) && (y != 0)
If x and y arrays have equal shape, the output shape is the same as these inputs.
Note: supports broadcasting if x and y have different shapes and are broadcastable.
Returns an array with values 1 where condition is satisfied, or value 0 otherwise.- Parameters:
x
- Input 1 (BOOL type)y
- Input 2 (BOOL type)- Returns:
- output INDArray with values 0 and 1 based on where the condition is satisfied (BOOL type)
-
asin
public INDArray asin(INDArray x)
Elementwise asin (arcsin, inverse sine) operation: out = arcsin(x)- Parameters:
x
- Input variable (NUMERIC type)- Returns:
- output Output variable (NUMERIC type)
-
asinh
public INDArray asinh(INDArray x)
Elementwise asinh (inverse hyperbolic sine) function: out = asinh(x)- Parameters:
x
- Input variable (NUMERIC type)- Returns:
- output Output variable (NUMERIC type)
-
asum
public INDArray asum(INDArray in, boolean keepDims, int... dimensions)
Absolute sum array reduction operation, optionally along specified dimensions: out = sum(abs(x))- Parameters:
in
- Input variable (NUMERIC type)keepDims
- Whether to keep the original dimensions or produce a shrunk array with less dimensionsdimensions
- Dimensions to reduce over. If dimensions are not specified, full array reduction is performed (Size: AtLeast(min=0))- Returns:
- output Reduced array of rank (input rank - num dimensions) (NUMERIC type)
-
asum
public INDArray asum(INDArray in, int... dimensions)
Absolute sum array reduction operation, optionally along specified dimensions: out = sum(abs(x))- Parameters:
in
- Input variable (NUMERIC type)dimensions
- Dimensions to reduce over. If dimensions are not specified, full array reduction is performed (Size: AtLeast(min=0))- Returns:
- output Reduced array of rank (input rank - num dimensions) (NUMERIC type)
-
atan
public INDArray atan(INDArray x)
Elementwise atan (arctangent, inverse tangent) operation: out = arctangent(x)- Parameters:
x
- Input variable (NUMERIC type)- Returns:
- output Output variable (NUMERIC type)
-
atan2
public INDArray atan2(INDArray y, INDArray x)
Elementwise atan (arctangent, inverse tangent) operation: out = atan2(x,y).
Similar to atan(y/x) but sigts of x and y are used to determine the location of the result- Parameters:
y
- Input Y variable (NUMERIC type)x
- Input X variable (NUMERIC type)- Returns:
- output Output variable (NUMERIC type)
-
atanh
public INDArray atanh(INDArray x)
Elementwise atanh (inverse hyperbolic tangent) function: out = atanh(x)- Parameters:
x
- Input variable (NUMERIC type)- Returns:
- output Output variable (NUMERIC type)
-
bitShift
public INDArray bitShift(INDArray x, INDArray shift)
Bit shift operation- Parameters:
x
- input (NUMERIC type)shift
- shift value (NUMERIC type)- Returns:
- output shifted output (NUMERIC type)
-
bitShiftRight
public INDArray bitShiftRight(INDArray x, INDArray shift)
Right bit shift operation- Parameters:
x
- Input tensor (NUMERIC type)shift
- shift argument (NUMERIC type)- Returns:
- output shifted output (NUMERIC type)
-
bitShiftRotl
public INDArray bitShiftRotl(INDArray x, INDArray shift)
Cyclic bit shift operation- Parameters:
x
- Input tensor (NUMERIC type)shift
- shift argy=ument (NUMERIC type)- Returns:
- output shifted output (NUMERIC type)
-
bitShiftRotr
public INDArray bitShiftRotr(INDArray x, INDArray shift)
Cyclic right shift operation- Parameters:
x
- Input tensor (NUMERIC type)shift
- Shift argument (NUMERIC type)- Returns:
- output Shifted output (NUMERIC type)
-
ceil
public INDArray ceil(INDArray x)
Element-wise ceiling function: out = ceil(x).
Rounds each value up to the nearest integer value (if not already an integer)- Parameters:
x
- Input variable (NUMERIC type)- Returns:
- output Output variable (NUMERIC type)
-
clipByNorm
public INDArray clipByNorm(INDArray x, double clipValue, int... dimensions)
Clipping by L2 norm, optionally along dimension(s)
if l2Norm(x,dimension) < clipValue, then input is returned unmodifed
Otherwise, out[i] = in[i] * clipValue / l2Norm(in, dimensions) where each value is clipped according
to the corresponding l2Norm along the specified dimensions- Parameters:
x
- Input variable (NUMERIC type)clipValue
- Clipping value (maximum l2 norm)dimensions
- Dimensions to reduce over. If dimensions are not specified, full array reduction is performed (Size: AtLeast(min=0))- Returns:
- output Output variable (NUMERIC type)
-
clipByValue
public INDArray clipByValue(INDArray x, double clipValueMin, double clipValueMax)
Element-wise clipping function:
out[i] = in[i] if in[i] >= clipValueMin and in[i] <= clipValueMax
out[i] = clipValueMin if in[i] < clipValueMin
out[i] = clipValueMax if in[i] > clipValueMax- Parameters:
x
- Input variable (NUMERIC type)clipValueMin
- Minimum value for clippingclipValueMax
- Maximum value for clipping- Returns:
- output Output variable (NUMERIC type)
-
confusionMatrix
public INDArray confusionMatrix(INDArray labels, INDArray pred, DataType dataType)
Compute the 2d confusion matrix of size [numClasses, numClasses] from a pair of labels and predictions, both of
which are represented as integer values. This version assumes the number of classes is 1 + max(max(labels), max(pred))
For example, if labels = [0, 1, 1] and predicted = [0, 2, 1] then output is:
[1, 0, 0]
[0, 1, 1]
[0, 0, 0]- Parameters:
labels
- Labels - 1D array of integer values representing label values (NUMERIC type)pred
- Predictions - 1D array of integer values representing predictions. Same length as labels (NUMERIC type)dataType
- Data type- Returns:
- output variable (2D, shape [numClasses, numClasses}) (NUMERIC type)
-
confusionMatrix
public INDArray confusionMatrix(INDArray labels, INDArray pred, int numClasses)
Compute the 2d confusion matrix of size [numClasses, numClasses] from a pair of labels and predictions, both of
which are represented as integer values.
For example, if labels = [0, 1, 1], predicted = [0, 2, 1], and numClasses=4 then output is:
[1, 0, 0, 0]
[0, 1, 1, 0]
[0, 0, 0, 0]
[0, 0, 0, 0]- Parameters:
labels
- Labels - 1D array of integer values representing label values (NUMERIC type)pred
- Predictions - 1D array of integer values representing predictions. Same length as labels (NUMERIC type)numClasses
- Number of classes- Returns:
- output variable (2D, shape [numClasses, numClasses}) (NUMERIC type)
-
confusionMatrix
public INDArray confusionMatrix(INDArray labels, INDArray pred, INDArray weights)
Compute the 2d confusion matrix of size [numClasses, numClasses] from a pair of labels and predictions, both of
which are represented as integer values. This version assumes the number of classes is 1 + max(max(labels), max(pred))
For example, if labels = [0, 1, 1], predicted = [0, 2, 1] and weights = [1, 2, 3]
[1, 0, 0]
[0, 3, 2]
[0, 0, 0]- Parameters:
labels
- Labels - 1D array of integer values representing label values (NUMERIC type)pred
- Predictions - 1D array of integer values representing predictions. Same length as labels (NUMERIC type)weights
- Weights - 1D array of values (may be real/decimal) representing the weight/contribution of each prediction. Must be same length as both labels and predictions arrays (NUMERIC type)- Returns:
- output variable (2D, shape [numClasses, numClasses}) (NUMERIC type)
-
confusionMatrix
public INDArray confusionMatrix(INDArray labels, INDArray pred, INDArray weights, int numClasses)
Compute the 2d confusion matrix of size [numClasses, numClasses] from a pair of labels and predictions, both of
which are represented as integer values.
For example, if labels = [0, 1, 1], predicted = [0, 2, 1], numClasses = 4, and weights = [1, 2, 3]
[1, 0, 0, 0]
[0, 3, 2, 0]
[0, 0, 0, 0]
[0, 0, 0, 0]- Parameters:
labels
- Labels - 1D array of integer values representing label values (NUMERIC type)pred
- Predictions - 1D array of integer values representing predictions. Same length as labels (NUMERIC type)weights
- Weights - 1D array of values (may be real/decimal) representing the weight/contribution of each prediction. Must be same length as both labels and predictions arrays (NUMERIC type)numClasses
-- Returns:
- output Output variable (2D, shape [numClasses, numClasses}) (NUMERIC type)
-
cos
public INDArray cos(INDArray x)
Elementwise cosine operation: out = cos(x)- Parameters:
x
- Input variable (NUMERIC type)- Returns:
- output Output variable (NUMERIC type)
-
cosh
public INDArray cosh(INDArray x)
Elementwise cosh (hyperbolic cosine) operation: out = cosh(x)- Parameters:
x
- Input variable (NUMERIC type)- Returns:
- output Output variable (NUMERIC type)
-
cosineDistance
public INDArray cosineDistance(INDArray x, INDArray y, boolean keepDims, boolean isComplex, int... dimensions)
Cosine distance reduction operation. The output contains the cosine distance for each
tensor/subset along the specified dimensions:
out = 1.0 - cosineSimilarity(x,y)- Parameters:
x
- Input variable x (NUMERIC type)y
- Input variable y (NUMERIC type)keepDims
- Whether to preserve original dimensions or notisComplex
- Depending on the implementation, such as distance calculations, this can determine whether all distance calculations for all points should be done.dimensions
- Dimensions to reduce over. If dimensions are not specified, full array reduction is performed (Size: AtLeast(min=0))- Returns:
- output Output variable (NUMERIC type)
-
cosineDistance
public INDArray cosineDistance(INDArray x, INDArray y, int... dimensions)
Cosine distance reduction operation. The output contains the cosine distance for each
tensor/subset along the specified dimensions:
out = 1.0 - cosineSimilarity(x,y)- Parameters:
x
- Input variable x (NUMERIC type)y
- Input variable y (NUMERIC type)dimensions
- Dimensions to reduce over. If dimensions are not specified, full array reduction is performed (Size: AtLeast(min=0))- Returns:
- output Output variable (NUMERIC type)
-
cosineSimilarity
public INDArray cosineSimilarity(INDArray x, INDArray y, boolean keepDims, boolean isComplex, int... dimensions)
Cosine similarity pairwise reduction operation. The output contains the cosine similarity for each tensor/subset
along the specified dimensions:
out = (sum_i x[i] * y[i]) / ( sqrt(sum_i x[i]^2) * sqrt(sum_i y[i]^2)- Parameters:
x
- Input variable x (NUMERIC type)y
- Input variable y (NUMERIC type)keepDims
- Whether to preserve original dimensions or notisComplex
- Depending on the implementation, such as distance calculations, this can determine whether all distance calculations for all points should be done.dimensions
- Dimensions to reduce over. If dimensions are not specified, full array reduction is performed (Size: AtLeast(min=0))- Returns:
- output Output variable (NUMERIC type)
-
cosineSimilarity
public INDArray cosineSimilarity(INDArray x, INDArray y, int... dimensions)
Cosine similarity pairwise reduction operation. The output contains the cosine similarity for each tensor/subset
along the specified dimensions:
out = (sum_i x[i] * y[i]) / ( sqrt(sum_i x[i]^2) * sqrt(sum_i y[i]^2)- Parameters:
x
- Input variable x (NUMERIC type)y
- Input variable y (NUMERIC type)dimensions
- Dimensions to reduce over. If dimensions are not specified, full array reduction is performed (Size: AtLeast(min=0))- Returns:
- output Output variable (NUMERIC type)
-
countNonZero
public INDArray countNonZero(INDArray in, boolean keepDims, int... dimensions)
Count non zero array reduction operation, optionally along specified dimensions: out = count(x != 0)- Parameters:
in
- Input variable (NUMERIC type)keepDims
- Whether to keep the original dimensions or produce a shrunk array with less dimensionsdimensions
- Dimensions to reduce over. If dimensions are not specified, full array reduction is performed (Size: AtLeast(min=0))- Returns:
- output Reduced array of rank (input rank - num dimensions) (NUMERIC type)
-
countNonZero
public INDArray countNonZero(INDArray in, int... dimensions)
Count non zero array reduction operation, optionally along specified dimensions: out = count(x != 0)- Parameters:
in
- Input variable (NUMERIC type)dimensions
- Dimensions to reduce over. If dimensions are not specified, full array reduction is performed (Size: AtLeast(min=0))- Returns:
- output Reduced array of rank (input rank - num dimensions) (NUMERIC type)
-
countZero
public INDArray countZero(INDArray in, boolean keepDims, int... dimensions)
Count zero array reduction operation, optionally along specified dimensions: out = count(x == 0)- Parameters:
in
- Input variable (NUMERIC type)keepDims
- Whether to keep the original dimensions or produce a shrunk array with less dimensionsdimensions
- Dimensions to reduce over. If dimensions are not specified, full array reduction is performed (Size: AtLeast(min=0))- Returns:
- output Reduced array of rank (input rank - num dimensions) (NUMERIC type)
-
countZero
public INDArray countZero(INDArray in, int... dimensions)
Count zero array reduction operation, optionally along specified dimensions: out = count(x == 0)- Parameters:
in
- Input variable (NUMERIC type)dimensions
- Dimensions to reduce over. If dimensions are not specified, full array reduction is performed (Size: AtLeast(min=0))- Returns:
- output Reduced array of rank (input rank - num dimensions) (NUMERIC type)
-
cross
public INDArray cross(INDArray a, INDArray b)
Returns the pair-wise cross product of equal size arrays a and b: a x b = ||a||x||b|| sin(theta).
Can take rank 1 or above inputs (of equal shapes), but note that the last dimension must have dimension 3- Parameters:
a
- First input (NUMERIC type)b
- Second input (NUMERIC type)- Returns:
- output Element-wise cross product (NUMERIC type)
-
cube
public INDArray cube(INDArray x)
Element-wise cube function: out = x^3- Parameters:
x
- Input variable (NUMERIC type)- Returns:
- output Output variable (NUMERIC type)
-
diag
public INDArray diag(INDArray x)
Returns an output variable with diagonal values equal to the specified values; off-diagonal values will be set to 0
For example, if input = [1,2,3], then output is given by:
[ 1, 0, 0]
[ 0, 2, 0]
[ 0, 0, 3]
Higher input ranks are also supported: if input has shape [a,...,R-1] then output[i,...,k,i,...,k] = input[i,...,k].
i.e., for input rank R, output has rank 2R- Parameters:
x
- Input variable (NUMERIC type)- Returns:
- output Output variable (NUMERIC type)
-
diagPart
public INDArray diagPart(INDArray x)
Extract the diagonal part from the input array.
If input is
[ 1, 0, 0]
[ 0, 2, 0]
[ 0, 0, 3]
then output is [1, 2, 3].
Supports higher dimensions: in general, out[i,...,k] = in[i,...,k,i,...,k]- Parameters:
x
- Input variable (NUMERIC type)- Returns:
- output Diagonal part of the input (NUMERIC type)
-
div
public INDArray div(INDArray x, INDArray y)
Pairwise division operation, out = x / y
Note: supports broadcasting if x and y have different shapes and are broadcastable.
For example, if X has shape [1,10] and Y has shape [5,10] then op(X,Y) has output shape [5,10]
Broadcast rules are the same as NumPy: https://docs.scipy.org/doc/numpy/user/basics.broadcasting.html- Parameters:
x
- Input variable (NUMERIC type)y
- Input variable (NUMERIC type)- Returns:
- output Output variable (NUMERIC type)
-
div
public INDArray div(INDArray x, double value)
Scalar division operation, out = in / scalar- Parameters:
x
- Input variable (NUMERIC type)value
- Scalar value for op- Returns:
- output Output variable (NUMERIC type)
-
entropy
public INDArray entropy(INDArray in, boolean keepDims, int... dimensions)
Entropy reduction: -sum(x * log(x))- Parameters:
in
- Input variable (NUMERIC type)keepDims
- Whether to keep the original dimensions or produce a shrunk array with less dimensionsdimensions
- Dimensions to reduce over. If dimensions are not specified, full array reduction is performed (Size: AtLeast(min=0))- Returns:
- output Reduced array of rank (input rank - num dimensions) (NUMERIC type)
-
entropy
public INDArray entropy(INDArray in, int... dimensions)
Entropy reduction: -sum(x * log(x))- Parameters:
in
- Input variable (NUMERIC type)dimensions
- Dimensions to reduce over. If dimensions are not specified, full array reduction is performed (Size: AtLeast(min=0))- Returns:
- output Reduced array of rank (input rank - num dimensions) (NUMERIC type)
-
erf
public INDArray erf(INDArray x)
Element-wise Gaussian error function - out = erf(in)- Parameters:
x
- Input variable (NUMERIC type)- Returns:
- output Output variable (NUMERIC type)
-
erfc
public INDArray erfc(INDArray x)
Element-wise complementary Gaussian error function - out = erfc(in) = 1 - erf(in)- Parameters:
x
- Input variable (NUMERIC type)- Returns:
- output Output variable (NUMERIC type)
-
euclideanDistance
public INDArray euclideanDistance(INDArray x, INDArray y, boolean keepDims, boolean isComplex, int... dimensions)
Euclidean distance (l2 norm, l2 distance) reduction operation. The output contains the Euclidean distance for each
tensor/subset along the specified dimensions:
out = sqrt( sum_i (x[i] - y[i])^2 )- Parameters:
x
- Input variable x (NUMERIC type)y
- Input variable y (NUMERIC type)keepDims
- Whether to preserve original dimensions or notisComplex
- Depending on the implementation, such as distance calculations, this can determine whether all distance calculations for all points should be done.dimensions
- Dimensions to reduce over. If dimensions are not specified, full array reduction is performed (Size: AtLeast(min=0))- Returns:
- output Output variable (NUMERIC type)
-
euclideanDistance
public INDArray euclideanDistance(INDArray x, INDArray y, int... dimensions)
Euclidean distance (l2 norm, l2 distance) reduction operation. The output contains the Euclidean distance for each
tensor/subset along the specified dimensions:
out = sqrt( sum_i (x[i] - y[i])^2 )- Parameters:
x
- Input variable x (NUMERIC type)y
- Input variable y (NUMERIC type)dimensions
- Dimensions to reduce over. If dimensions are not specified, full array reduction is performed (Size: AtLeast(min=0))- Returns:
- output Output variable (NUMERIC type)
-
exp
public INDArray exp(INDArray x)
Elementwise exponent function: out = exp(x) = 2.71828...^x- Parameters:
x
- Input variable (NUMERIC type)- Returns:
- output Output variable (NUMERIC type)
-
expm1
public INDArray expm1(INDArray x)
Elementwise 1.0 - exponent function: out = 1.0 - exp(x) = 1.0 - 2.71828...^x- Parameters:
x
- Input variable (NUMERIC type)- Returns:
- output Output variable (NUMERIC type)
-
eye
public INDArray eye(int rows)
Generate an identity matrix with the specified number of rows and columns.- Parameters:
rows
- Number of rows- Returns:
- output Identity matrix (NUMERIC type)
-
eye
public INDArray eye(int rows, int cols)
As per eye(String, int, int, DataType) but with the default datatype, Eye.DEFAULT_DTYPE- Parameters:
rows
- Number of rowscols
- Number of columns- Returns:
- output (NUMERIC type)
-
eye
public INDArray eye(int rows, int cols, DataType dataType, int... dimensions)
Generate an identity matrix with the specified number of rows and columns
Example:
INDArray eye = eye(3,2)<br> eye:<br> [ 1, 0]<br> [ 0, 1]<br> [ 0, 0]
- Parameters:
rows
- Number of rowscols
- Number of columnsdataType
- Data typedimensions
- (Size: AtLeast(min=0))- Returns:
- output Identity matrix (NUMERIC type)
-
eye
public INDArray eye(INDArray rows, INDArray cols)
As per eye(int, int) bit with the number of rows/columns specified as scalar INDArrays- Parameters:
rows
- Number of rows (INT type)cols
- Number of columns (INT type)- Returns:
- output Identity matrix (NUMERIC type)
-
eye
public INDArray eye(INDArray rows)
As per eye(String, int) but with the number of rows specified as a scalar INDArray- Parameters:
rows
- Number of rows (INT type)- Returns:
- output SDVaribable identity matrix (NUMERIC type)
-
firstIndex
public INDArray firstIndex(INDArray in, Condition condition, int... dimensions)
First index reduction operation.
Returns a variable that contains the index of the first element that matches the specified condition (for each
slice along the specified dimensions)
Note that if keepDims = true, the output variable has the same rank as the input variable,
with the reduced dimensions having size 1. This can be useful for later broadcast operations (such as subtracting
the mean along a dimension).
Example: if input has shape [a,b,c] and dimensions=[1] then output has shape:
keepDims = true: [a,1,c]
keepDims = false: [a,c]- Parameters:
in
- Input variable (NUMERIC type)condition
- Condition to check on input variabledimensions
- Dimensions to reduce over. If dimensions are not specified, full array reduction is performed (Size: AtLeast(min=1))- Returns:
- output Reduced array of rank (input rank - num dimensions) (NUMERIC type)
-
firstIndex
public INDArray firstIndex(INDArray in, Condition condition, boolean keepDims, int... dimensions)
First index reduction operation.
Returns a variable that contains the index of the first element that matches the specified condition (for each
slice along the specified dimensions)
Note that if keepDims = true, the output variable has the same rank as the input variable,
with the reduced dimensions having size 1. This can be useful for later broadcast operations (such as subtracting
the mean along a dimension).
Example: if input has shape [a,b,c] and dimensions=[1] then output has shape:
keepDims = true: [a,1,c]
keepDims = false: [a,c]- Parameters:
in
- Input variable (NUMERIC type)condition
- Condition to check on input variablekeepDims
- If true: keep the dimensions that are reduced on (as length 1). False: remove the reduction dimensionsdimensions
- Dimensions to reduce over. If dimensions are not specified, full array reduction is performed (Size: AtLeast(min=1))- Returns:
- output Reduced array of rank (input rank - num dimensions) (NUMERIC type)
-
floor
public INDArray floor(INDArray x)
Element-wise floor function: out = floor(x).
Rounds each value down to the nearest integer value (if not already an integer)- Parameters:
x
- Input variable (NUMERIC type)- Returns:
- output Output variable (NUMERIC type)
-
floorDiv
public INDArray floorDiv(INDArray x, INDArray y)
Pairwise floor division operation, out = floor(x / y)
Note: supports broadcasting if x and y have different shapes and are broadcastable.
For example, if X has shape [1,10] and Y has shape [5,10] then op(X,Y) has output shape [5,10]
Broadcast rules are the same as NumPy: https://docs.scipy.org/doc/numpy/user/basics.broadcasting.html- Parameters:
x
- Input variable (NUMERIC type)y
- Input variable (NUMERIC type)- Returns:
- output Output variable (NUMERIC type)
-
floorMod
public INDArray floorMod(INDArray x, INDArray y)
Pairwise Modulus division operation
Note: supports broadcasting if x and y have different shapes and are broadcastable.
For example, if X has shape [1,10] and Y has shape [5,10] then op(X,Y) has output shape [5,10]
Broadcast rules are the same as NumPy: https://docs.scipy.org/doc/numpy/user/basics.broadcasting.html- Parameters:
x
- Input variable (NUMERIC type)y
- Input variable (NUMERIC type)- Returns:
- output Output variable (NUMERIC type)
-
floorMod
public INDArray floorMod(INDArray x, double value)
Scalar floor modulus operation- Parameters:
x
- Input variable (NUMERIC type)value
- Scalar value for op- Returns:
- output Output variable (NUMERIC type)
-
hammingDistance
public INDArray hammingDistance(INDArray x, INDArray y, boolean keepDims, boolean isComplex, int... dimensions)
Hamming distance reduction operation. The output contains the cosine distance for each
tensor/subset along the specified dimensions:
out = count( x[i] != y[i] )- Parameters:
x
- Input variable x (NUMERIC type)y
- Input variable y (NUMERIC type)keepDims
- Whether to preserve original dimensions or notisComplex
- Depending on the implementation, such as distance calculations, this can determine whether all distance calculations for all points should be done.dimensions
- Dimensions to reduce over. If dimensions are not specified, full array reduction is performed (Size: AtLeast(min=0))- Returns:
- output Output variable (NUMERIC type)
-
hammingDistance
public INDArray hammingDistance(INDArray x, INDArray y, int... dimensions)
Hamming distance reduction operation. The output contains the cosine distance for each
tensor/subset along the specified dimensions:
out = count( x[i] != y[i] )- Parameters:
x
- Input variable x (NUMERIC type)y
- Input variable y (NUMERIC type)dimensions
- Dimensions to reduce over. If dimensions are not specified, full array reduction is performed (Size: AtLeast(min=0))- Returns:
- output Output variable (NUMERIC type)
-
iamax
public INDArray iamax(INDArray in, int... dimensions)
Index of the max absolute value: argmax(abs(in))
see argmax(String, INDArray, boolean, int...)- Parameters:
in
- Input variable (NUMERIC type)dimensions
- Dimensions to reduce over. If dimensions are not specified, full array reduction is performed (Size: AtLeast(min=1))- Returns:
- output Reduced array of rank (input rank - num dimensions) (NUMERIC type)
-
iamax
public INDArray iamax(INDArray in, boolean keepDims, int... dimensions)
Index of the max absolute value: argmax(abs(in))
see argmax(String, INDArray, boolean, int...)- Parameters:
in
- Input variable (NUMERIC type)keepDims
- If true: keep the dimensions that are reduced on (as length 1). False: remove the reduction dimensionsdimensions
- Dimensions to reduce over. If dimensions are not specified, full array reduction is performed (Size: AtLeast(min=1))- Returns:
- output Reduced array of rank (input rank - num dimensions) (NUMERIC type)
-
iamin
public INDArray iamin(INDArray in, int... dimensions)
Index of the min absolute value: argmin(abs(in))
see argmin(String, INDArray, boolean, int...)- Parameters:
in
- Input variable (NUMERIC type)dimensions
- Dimensions to reduce over. If dimensions are not specified, full array reduction is performed (Size: AtLeast(min=1))- Returns:
- output Reduced array of rank (input rank - num dimensions) (NUMERIC type)
-
iamin
public INDArray iamin(INDArray in, boolean keepDims, int... dimensions)
Index of the min absolute value: argmin(abs(in))
see argmin(String, INDArray, boolean, int...)- Parameters:
in
- Input variable (NUMERIC type)keepDims
- If true: keep the dimensions that are reduced on (as length 1). False: remove the reduction dimensionsdimensions
- Dimensions to reduce over. If dimensions are not specified, full array reduction is performed (Size: AtLeast(min=1))- Returns:
- output Reduced array of rank (input rank - num dimensions) (NUMERIC type)
-
isFinite
public INDArray isFinite(INDArray x)
Is finite operation: elementwise isFinite(x)
Returns an array with the same shape/size as the input, with values 1 where condition is satisfied, or
value 0 otherwise- Parameters:
x
- Input variable (NUMERIC type)- Returns:
- output Output variable (NUMERIC type)
-
isInfinite
public INDArray isInfinite(INDArray x)
Is infinite operation: elementwise isInfinite(x)
Returns an array with the same shape/size as the input, with values 1 where condition is satisfied, or
value 0 otherwise- Parameters:
x
- Input variable (NUMERIC type)- Returns:
- output Output variable (NUMERIC type)
-
isMax
public INDArray isMax(INDArray x)
Is maximum operation: elementwise x == max(x)
Returns an array with the same shape/size as the input, with values 1 where condition is satisfied, or
value 0 otherwise- Parameters:
x
- Input variable (NUMERIC type)- Returns:
- output Output variable (NUMERIC type)
-
isNaN
public INDArray isNaN(INDArray x)
Is Not a Number operation: elementwise isNaN(x)
Returns an array with the same shape/size as the input, with values 1 where condition is satisfied, or
value 0 otherwise- Parameters:
x
- Input variable (NUMERIC type)- Returns:
- output Output variable (NUMERIC type)
-
isNonDecreasing
public INDArray isNonDecreasing(INDArray x)
Is the array non decreasing?
An array is non-decreasing if for every valid i, x[i] <= x[i+1]. For Rank 2+ arrays, values are compared
in 'c' (row major) order- Parameters:
x
- Input variable (NUMERIC type)- Returns:
- output Scalar variable with value 1 if non-decreasing, or 0 otherwise (NUMERIC type)
-
isStrictlyIncreasing
public INDArray isStrictlyIncreasing(INDArray x)
Is the array strictly increasing?
An array is strictly increasing if for every valid i, x[i] < x[i+1]. For Rank 2+ arrays, values are compared
in 'c' (row major) order- Parameters:
x
- Input variable (NUMERIC type)- Returns:
- output Scalar variable with value 1 if strictly increasing, or 0 otherwise (NUMERIC type)
-
jaccardDistance
public INDArray jaccardDistance(INDArray x, INDArray y, boolean keepDims, boolean isComplex, int... dimensions)
Jaccard similarity reduction operation. The output contains the Jaccard distance for each
tensor along the specified dimensions.- Parameters:
x
- Input variable x (NUMERIC type)y
- Input variable y (NUMERIC type)keepDims
- Whether to preserve original dimensions or notisComplex
- Depending on the implementation, such as distance calculations, this can determine whether all distance calculations for all points should be done.dimensions
- Dimensions to reduce over. If dimensions are not specified, full array reduction is performed (Size: AtLeast(min=0))- Returns:
- output Output variable (NUMERIC type)
-
jaccardDistance
public INDArray jaccardDistance(INDArray x, INDArray y, int... dimensions)
Jaccard similarity reduction operation. The output contains the Jaccard distance for each
tensor along the specified dimensions.- Parameters:
x
- Input variable x (NUMERIC type)y
- Input variable y (NUMERIC type)dimensions
- Dimensions to reduce over. If dimensions are not specified, full array reduction is performed (Size: AtLeast(min=0))- Returns:
- output Output variable (NUMERIC type)
-
lastIndex
public INDArray lastIndex(INDArray in, Condition condition, int... dimensions)
Last index reduction operation.
Returns a variable that contains the index of the last element that matches the specified condition (for each
slice along the specified dimensions)
Note that if keepDims = true, the output variable has the same rank as the input variable,
with the reduced dimensions having size 1. This can be useful for later broadcast operations (such as subtracting
the mean along a dimension).
Example: if input has shape [a,b,c] and dimensions=[1] then output has shape:
keepDims = true: [a,1,c]
keepDims = false: [a,c]- Parameters:
in
- Input variable (NUMERIC type)condition
- Condition to check on input variabledimensions
- Dimensions to reduce over. If dimensions are not specified, full array reduction is performed (Size: AtLeast(min=1))- Returns:
- output Reduced array of rank (input rank - num dimensions) (NUMERIC type)
-
lastIndex
public INDArray lastIndex(INDArray in, Condition condition, boolean keepDims, int... dimensions)
Last index reduction operation.
Returns a variable that contains the index of the last element that matches the specified condition (for each
slice along the specified dimensions)
Note that if keepDims = true, the output variable has the same rank as the input variable,
with the reduced dimensions having size 1. This can be useful for later broadcast operations (such as subtracting
the mean along a dimension).
Example: if input has shape [a,b,c] and dimensions=[1] then output has shape:
keepDims = true: [a,1,c]
keepDims = false: [a,c]- Parameters:
in
- Input variable (NUMERIC type)condition
- Condition to check on input variablekeepDims
- If true: keep the dimensions that are reduced on (as length 1). False: remove the reduction dimensionsdimensions
- Dimensions to reduce over. If dimensions are not specified, full array reduction is performed (Size: AtLeast(min=1))- Returns:
- output Reduced array of rank (input rank - num dimensions) (NUMERIC type)
-
listDiff
public INDArray[] listDiff(INDArray x, INDArray y)
Calculates difference between inputs X and Y.- Parameters:
x
- Input variable X (NUMERIC type)y
- Input variable Y (NUMERIC type)
-
log
public INDArray log(INDArray x)
Element-wise logarithm function (base e - natural logarithm): out = log(x)- Parameters:
x
- Input variable (NUMERIC type)- Returns:
- output Output variable (NUMERIC type)
-
log
public INDArray log(INDArray x, double base)
Element-wise logarithm function (with specified base): out = log_{base}(x)- Parameters:
x
- Input variable (NUMERIC type)base
- Logarithm base- Returns:
- output Output variable (NUMERIC type)
-
log1p
public INDArray log1p(INDArray x)
Elementwise natural logarithm function: out = log_e (1 + x)- Parameters:
x
- Input variable (NUMERIC type)- Returns:
- output Output variable (NUMERIC type)
-
logEntropy
public INDArray logEntropy(INDArray in, boolean keepDims, int... dimensions)
Log entropy reduction: log(-sum(x * log(x)))- Parameters:
in
- Input variable (NUMERIC type)keepDims
- Whether to keep the original dimensions or produce a shrunk array with less dimensionsdimensions
- Dimensions to reduce over. If dimensions are not specified, full array reduction is performed (Size: AtLeast(min=0))- Returns:
- output Reduced array of rank (input rank - num dimensions) (NUMERIC type)
-
logEntropy
public INDArray logEntropy(INDArray in, int... dimensions)
Log entropy reduction: log(-sum(x * log(x)))- Parameters:
in
- Input variable (NUMERIC type)dimensions
- Dimensions to reduce over. If dimensions are not specified, full array reduction is performed (Size: AtLeast(min=0))- Returns:
- output Reduced array of rank (input rank - num dimensions) (NUMERIC type)
-
logSumExp
public INDArray logSumExp(INDArray input, int... dimensions)
Log-sum-exp reduction (optionally along dimension).
Computes log(sum(exp(x))- Parameters:
input
- Input variable (NUMERIC type)dimensions
- Optional dimensions to reduce along (Size: AtLeast(min=0))- Returns:
- output Output variable (NUMERIC type)
-
manhattanDistance
public INDArray manhattanDistance(INDArray x, INDArray y, boolean keepDims, boolean isComplex, int... dimensions)
Manhattan distance (l1 norm, l1 distance) reduction operation. The output contains the Manhattan distance for each
tensor/subset along the specified dimensions:
out = sum_i abs(x[i]-y[i])- Parameters:
x
- Input variable x (NUMERIC type)y
- Input variable y (NUMERIC type)keepDims
- Whether to preserve original dimensions or notisComplex
- Depending on the implementation, such as distance calculations, this can determine whether all distance calculations for all points should be done.dimensions
- Dimensions to reduce over. If dimensions are not specified, full array reduction is performed (Size: AtLeast(min=0))- Returns:
- output Output variable (NUMERIC type)
-
manhattanDistance
public INDArray manhattanDistance(INDArray x, INDArray y, int... dimensions)
Manhattan distance (l1 norm, l1 distance) reduction operation. The output contains the Manhattan distance for each
tensor/subset along the specified dimensions:
out = sum_i abs(x[i]-y[i])- Parameters:
x
- Input variable x (NUMERIC type)y
- Input variable y (NUMERIC type)dimensions
- Dimensions to reduce over. If dimensions are not specified, full array reduction is performed (Size: AtLeast(min=0))- Returns:
- output Output variable (NUMERIC type)
-
matrixDeterminant
public INDArray matrixDeterminant(INDArray in)
Matrix determinant op. For 2D input, this returns the standard matrix determinant.
For higher dimensional input with shape [..., m, m] the matrix determinant is returned for each
shape [m,m] sub-matrix.- Parameters:
in
- Input (NUMERIC type)- Returns:
- output Matrix determinant variable (NUMERIC type)
-
matrixInverse
public INDArray matrixInverse(INDArray in)
Matrix inverse op. For 2D input, this returns the standard matrix inverse.
For higher dimensional input with shape [..., m, m] the matrix inverse is returned for each
shape [m,m] sub-matrix.- Parameters:
in
- Input (NUMERIC type)- Returns:
- output Matrix inverse variable (NUMERIC type)
-
max
public INDArray max(INDArray x, INDArray y)
Pairwise max operation, out = max(x, y)
Note: supports broadcasting if x and y have different shapes and are broadcastable.
For example, if X has shape [1,10] and Y has shape [5,10] then op(X,Y) has output shape [5,10]
Broadcast rules are the same as NumPy: https://docs.scipy.org/doc/numpy/user/basics.broadcasting.html- Parameters:
x
- First input variable, x (NUMERIC type)y
- Second input variable, y (NUMERIC type)- Returns:
- out Output (NUMERIC type)
-
mean
public INDArray mean(INDArray in, boolean keepDims, int... dimensions)
Mean array reduction operation, optionally along specified dimensions: out = mean(abs(x))- Parameters:
in
- Input variable (NUMERIC type)keepDims
- Whether to keep the original dimensions or produce a shrunk array with less dimensionsdimensions
- Dimensions to reduce over. If dimensions are not specified, full array reduction is performed (Size: AtLeast(min=0))- Returns:
- output Reduced array of rank (input rank - num dimensions) (NUMERIC type)
-
mean
public INDArray mean(INDArray in, int... dimensions)
Mean array reduction operation, optionally along specified dimensions: out = mean(abs(x))- Parameters:
in
- Input variable (NUMERIC type)dimensions
- Dimensions to reduce over. If dimensions are not specified, full array reduction is performed (Size: AtLeast(min=0))- Returns:
- output Reduced array of rank (input rank - num dimensions) (NUMERIC type)
-
mean
public INDArray mean(INDArray in, INDArray dimensions, boolean keepDims)
Mean array reduction operation, optionally along specified dimensions: out = mean(abs(x))- Parameters:
in
- Input variable (NUMERIC type)dimensions
- Dimensions to reduce along (NUMERIC type)keepDims
- Whether to keep the original dimensions or produce a shrunk array with less dimensions- Returns:
- output Reduced array of rank (input rank - num dimensions) (NUMERIC type)
-
mean
public INDArray mean(INDArray in, INDArray dimensions)
Mean array reduction operation, optionally along specified dimensions: out = mean(abs(x))- Parameters:
in
- Input variable (NUMERIC type)dimensions
- Dimensions to reduce along (NUMERIC type)- Returns:
- output Reduced array of rank (input rank - num dimensions) (NUMERIC type)
-
mergeAdd
public INDArray mergeAdd(INDArray... inputs)
Merge add function: merges an arbitrary number of equal shaped arrays using element-wise addition:
out = sum_i in[i]- Parameters:
inputs
- Input variables (NUMERIC type)- Returns:
- output Output variable (NUMERIC type)
-
mergeAvg
public INDArray mergeAvg(INDArray... inputs)
Merge average function: merges an arbitrary number of equal shaped arrays using element-wise mean operation:
out = mean_i in[i]- Parameters:
inputs
- Input variables (NUMERIC type)- Returns:
- output Output variable (NUMERIC type)
-
mergeMax
public INDArray mergeMax(INDArray... inputs)
Merge max function: merges an arbitrary number of equal shaped arrays using element-wise maximum operation:
out = max_i in[i]- Parameters:
inputs
- Input variables (NUMERIC type)- Returns:
- output Output variable (NUMERIC type)
-
meshgrid
public INDArray[] meshgrid(INDArray[] inputs, boolean cartesian)
Broadcasts parameters for evaluation on an N-D grid.- Parameters:
inputs
- (NUMERIC type)cartesian
-
-
min
public INDArray min(INDArray x, INDArray y)
Pairwise max operation, out = min(x, y)
Note: supports broadcasting if x and y have different shapes and are broadcastable.
For example, if X has shape [1,10] and Y has shape [5,10] then op(X,Y) has output shape [5,10]
Broadcast rules are the same as NumPy: https://docs.scipy.org/doc/numpy/user/basics.broadcasting.html- Parameters:
x
- First input variable, x (NUMERIC type)y
- Second input variable, y (NUMERIC type)- Returns:
- out Output (NUMERIC type)
-
mod
public INDArray mod(INDArray x, INDArray y)
Pairwise modulus (remainder) operation, out = x % y
Note: supports broadcasting if x and y have different shapes and are broadcastable.
For example, if X has shape [1,10] and Y has shape [5,10] then op(X,Y) has output shape [5,10]
Broadcast rules are the same as NumPy: https://docs.scipy.org/doc/numpy/user/basics.broadcasting.html- Parameters:
x
- Input variable (NUMERIC type)y
- Input variable (NUMERIC type)- Returns:
- output Output variable (NUMERIC type)
-
moments
public INDArray[] moments(INDArray input, int[] axes, boolean keepDims)
Calculate the mean and (population) variance for the input variable, for the specified axis- Parameters:
input
- Input to calculate moments for (NUMERIC type)axes
- Dimensions to perform calculation over (Size: AtLeast(min=0))keepDims
- Whether to keep dimensions during reduction or not.
-
moments
public INDArray[] moments(INDArray input, INDArray axes, boolean keepDims)
Calculate the mean and (population) variance for the input variable, for the specified axis- Parameters:
input
- Input to calculate moments for (NUMERIC type)axes
- Dimensions to perform calculation over (NUMERIC type)keepDims
- Whether to keep dimensions during reduction or not.
-
mul
public INDArray mul(INDArray x, INDArray y)
Pairwise multiplication operation, out = x * y
Note: supports broadcasting if x and y have different shapes and are broadcastable.
For example, if X has shape [1,10] and Y has shape [5,10] then op(X,Y) has output shape [5,10]
Broadcast rules are the same as NumPy: https://docs.scipy.org/doc/numpy/user/basics.broadcasting.html- Parameters:
x
- Input variable (NUMERIC type)y
- Input variable (NUMERIC type)- Returns:
- output Output variable (NUMERIC type)
-
mul
public INDArray mul(INDArray x, double value)
Scalar multiplication operation, out = in * scalar- Parameters:
x
- Input variable (NUMERIC type)value
- Scalar value for op- Returns:
- output Output variable (NUMERIC type)
-
neg
public INDArray neg(INDArray x)
Elementwise negative operation: out = -x- Parameters:
x
- Input variable (NUMERIC type)- Returns:
- output Output variable (NUMERIC type)
-
norm1
public INDArray norm1(INDArray in, boolean keepDims, int... dimensions)
Mean array reduction operation, optionally along specified dimensions: out = mean(abs(x))- Parameters:
in
- Input variable (NUMERIC type)keepDims
- Whether to keep the original dimensions or produce a shrunk array with less dimensionsdimensions
- Dimensions to reduce over. If dimensions are not specified, full array reduction is performed (Size: AtLeast(min=0))- Returns:
- output Reduced array of rank (input rank - num dimensions) (NUMERIC type)
-
norm1
public INDArray norm1(INDArray in, int... dimensions)
Mean array reduction operation, optionally along specified dimensions: out = mean(abs(x))- Parameters:
in
- Input variable (NUMERIC type)dimensions
- Dimensions to reduce over. If dimensions are not specified, full array reduction is performed (Size: AtLeast(min=0))- Returns:
- output Reduced array of rank (input rank - num dimensions) (NUMERIC type)
-
norm1
public INDArray norm1(INDArray in, INDArray dimensions, boolean keepDims)
Sum of absolute differences.- Parameters:
in
- Input variable (NUMERIC type)dimensions
- Dimensions to reduce along (NUMERIC type)keepDims
- Whether to keep the original dimensions or produce a shrunk array with less dimensions- Returns:
- output Reduced array of rank (input rank - num dimensions) (NUMERIC type)
-
norm1
public INDArray norm1(INDArray in, INDArray dimensions)
Sum of absolute differences.- Parameters:
in
- Input variable (NUMERIC type)dimensions
- Dimensions to reduce along (NUMERIC type)- Returns:
- output Reduced array of rank (input rank - num dimensions) (NUMERIC type)
-
norm2
public INDArray norm2(INDArray in, boolean keepDims, int... dimensions)
Euclidean norm: euclidean distance of a vector from the origin- Parameters:
in
- Input variable (NUMERIC type)keepDims
- Whether to keep the original dimensions or produce a shrunk array with less dimensionsdimensions
- Dimensions to reduce over. If dimensions are not specified, full array reduction is performed (Size: AtLeast(min=0))- Returns:
- output Reduced array of rank (input rank - num dimensions) (NUMERIC type)
-
norm2
public INDArray norm2(INDArray in, int... dimensions)
Euclidean norm: euclidean distance of a vector from the origin- Parameters:
in
- Input variable (NUMERIC type)dimensions
- Dimensions to reduce over. If dimensions are not specified, full array reduction is performed (Size: AtLeast(min=0))- Returns:
- output Reduced array of rank (input rank - num dimensions) (NUMERIC type)
-
norm2
public INDArray norm2(INDArray in, INDArray dimensions, boolean keepDims)
Euclidean norm: euclidean distance of a vector from the origin- Parameters:
in
- Input variable (NUMERIC type)dimensions
- Dimensions to reduce along (NUMERIC type)keepDims
- Whether to keep the original dimensions or produce a shrunk array with less dimensions- Returns:
- output Reduced array of rank (input rank - num dimensions) (NUMERIC type)
-
norm2
public INDArray norm2(INDArray in, INDArray dimensions)
Euclidean norm: euclidean distance of a vector from the origin- Parameters:
in
- Input variable (NUMERIC type)dimensions
- Dimensions to reduce along (NUMERIC type)- Returns:
- output Reduced array of rank (input rank - num dimensions) (NUMERIC type)
-
normMax
public INDArray normMax(INDArray in, boolean keepDims, int... dimensions)
Differences between max absolute value- Parameters:
in
- Input variable (NUMERIC type)keepDims
- Whether to keep the original dimensions or produce a shrunk array with less dimensionsdimensions
- Dimensions to reduce over. If dimensions are not specified, full array reduction is performed (Size: AtLeast(min=0))- Returns:
- output Reduced array of rank (input rank - num dimensions) (NUMERIC type)
-
normMax
public INDArray normMax(INDArray in, int... dimensions)
Differences between max absolute value- Parameters:
in
- Input variable (NUMERIC type)dimensions
- Dimensions to reduce over. If dimensions are not specified, full array reduction is performed (Size: AtLeast(min=0))- Returns:
- output Reduced array of rank (input rank - num dimensions) (NUMERIC type)
-
normMax
public INDArray normMax(INDArray in, INDArray dimensions, boolean keepDims)
Differences between max absolute value- Parameters:
in
- Input variable (NUMERIC type)dimensions
- Dimensions to reduce along (NUMERIC type)keepDims
- Whether to keep the original dimensions or produce a shrunk array with less dimensions- Returns:
- output Reduced array of rank (input rank - num dimensions) (NUMERIC type)
-
normMax
public INDArray normMax(INDArray in, INDArray dimensions)
Differences between max absolute value- Parameters:
in
- Input variable (NUMERIC type)dimensions
- Dimensions to reduce along (NUMERIC type)- Returns:
- output Reduced array of rank (input rank - num dimensions) (NUMERIC type)
-
normalizeMoments
public INDArray[] normalizeMoments(INDArray counts, INDArray means, INDArray variances, double shift)
Calculate the mean and variance from the sufficient statistics- Parameters:
counts
- Rank 0 (scalar) value with the total number of values used to calculate the sufficient statistics (NUMERIC type)means
- Mean-value sufficient statistics: this is the SUM of all data values (NUMERIC type)variances
- Variaance sufficient statistics: this is the squared sum of all data values (NUMERIC type)shift
- Shift value, possibly 0, used when calculating the sufficient statistics (for numerical stability)
-
or
public INDArray or(INDArray x, INDArray y)
Boolean OR operation: elementwise (x != 0) || (y != 0)
If x and y arrays have equal shape, the output shape is the same as these inputs.
Note: supports broadcasting if x and y have different shapes and are broadcastable.
Returns an array with values 1 where condition is satisfied, or value 0 otherwise.- Parameters:
x
- Input 1 (BOOL type)y
- Input 2 (BOOL type)- Returns:
- output INDArray with values 0 and 1 based on where the condition is satisfied (BOOL type)
-
pow
public INDArray pow(INDArray x, double value)
Element-wise power function: out = x^value- Parameters:
x
- Input variable (NUMERIC type)value
- Scalar value for op- Returns:
- output Output variable (NUMERIC type)
-
pow
public INDArray pow(INDArray x, INDArray y)
Element-wise (broadcastable) power function: out = x[i]^y[i]- Parameters:
x
- Input variable (NUMERIC type)y
- Power (NUMERIC type)- Returns:
- output Output variable (NUMERIC type)
-
prod
public INDArray prod(INDArray in, boolean keepDims, int... dimensions)
The max of an array along each dimension- Parameters:
in
- Input variable (NUMERIC type)keepDims
- Whether to keep the original dimensions or produce a shrunk array with less dimensionsdimensions
- Dimensions to reduce over. If dimensions are not specified, full array reduction is performed (Size: AtLeast(min=0))- Returns:
- output Reduced array of rank (input rank - num dimensions) (NUMERIC type)
-
prod
public INDArray prod(INDArray in, int... dimensions)
The max of an array along each dimension- Parameters:
in
- Input variable (NUMERIC type)dimensions
- Dimensions to reduce over. If dimensions are not specified, full array reduction is performed (Size: AtLeast(min=0))- Returns:
- output Reduced array of rank (input rank - num dimensions) (NUMERIC type)
-
prod
public INDArray prod(INDArray in, INDArray dimensions, boolean keepDims)
The product of an array long each dimension- Parameters:
in
- Input variable (NUMERIC type)dimensions
- Dimensions to reduce along (NUMERIC type)keepDims
- Whether to keep the original dimensions or produce a shrunk array with less dimensions- Returns:
- output Reduced array of rank (input rank - num dimensions) (NUMERIC type)
-
prod
public INDArray prod(INDArray in, INDArray dimensions)
The product of an array long each dimension- Parameters:
in
- Input variable (NUMERIC type)dimensions
- Dimensions to reduce along (NUMERIC type)- Returns:
- output Reduced array of rank (input rank - num dimensions) (NUMERIC type)
-
rationalTanh
public INDArray rationalTanh(INDArray x)
Rational Tanh Approximation elementwise function, as described in the paper:
Compact Convolutional Neural Network Cascade for Face Detection
This is a faster Tanh approximation- Parameters:
x
- Input variable (NUMERIC type)- Returns:
- output Output variable (NUMERIC type)
-
rdiv
public INDArray rdiv(INDArray x, INDArray y)
Pairwise reverse division operation, out = y / x
Note: supports broadcasting if x and y have different shapes and are broadcastable.
For example, if X has shape [1,10] and Y has shape [5,10] then op(X,Y) has output shape [5,10]
Broadcast rules are the same as NumPy: https://docs.scipy.org/doc/numpy/user/basics.broadcasting.html- Parameters:
x
- Input variable (NUMERIC type)y
- Input variable (NUMERIC type)- Returns:
- output Output variable (NUMERIC type)
-
rdiv
public INDArray rdiv(INDArray x, double value)
Scalar reverse division operation, out = scalar / in- Parameters:
x
- Input variable (NUMERIC type)value
- Scalar value for op- Returns:
- output Output variable (NUMERIC type)
-
reciprocal
public INDArray reciprocal(INDArray x)
Element-wise reciprocal (inverse) function: out[i] = 1 / in[i]- Parameters:
x
- Input variable (NUMERIC type)- Returns:
- output Output variable (NUMERIC type)
-
rectifiedTanh
public INDArray rectifiedTanh(INDArray x)
Rectified tanh operation: max(0, tanh(in))- Parameters:
x
- Input variable (NUMERIC type)- Returns:
- output Output variable (NUMERIC type)
-
reduceAMax
public INDArray reduceAMax(INDArray in, boolean keepDims, int... dimensions)
Absolute max array reduction operation, optionally along specified dimensions: out = max(abs(x))- Parameters:
in
- Input variable (NUMERIC type)keepDims
- Whether to keep the original dimensions or produce a shrunk array with less dimensionsdimensions
- Dimensions to reduce over. If dimensions are not specified, full array reduction is performed (Size: AtLeast(min=0))- Returns:
- output Reduced array of rank (input rank - num dimensions) (NUMERIC type)
-
reduceAMax
public INDArray reduceAMax(INDArray in, int... dimensions)
Absolute max array reduction operation, optionally along specified dimensions: out = max(abs(x))- Parameters:
in
- Input variable (NUMERIC type)dimensions
- Dimensions to reduce over. If dimensions are not specified, full array reduction is performed (Size: AtLeast(min=0))- Returns:
- output Reduced array of rank (input rank - num dimensions) (NUMERIC type)
-
reduceAMax
public INDArray reduceAMax(INDArray in, INDArray dimensions, boolean keepDims)
Absolute max array reduction operation, optionally along specified dimensions: out = max(abs(x))- Parameters:
in
- Input variable (NUMERIC type)dimensions
- Dimensions to reduce along (NUMERIC type)keepDims
- Whether to keep the original dimensions or produce a shrunk array with less dimensions- Returns:
- output Reduced array of rank (input rank - num dimensions) (NUMERIC type)
-
reduceAMax
public INDArray reduceAMax(INDArray in, INDArray dimensions)
Absolute max array reduction operation, optionally along specified dimensions: out = max(abs(x))- Parameters:
in
- Input variable (NUMERIC type)dimensions
- Dimensions to reduce along (NUMERIC type)- Returns:
- output Reduced array of rank (input rank - num dimensions) (NUMERIC type)
-
reduceAmean
public INDArray reduceAmean(INDArray in, boolean keepDims, int... dimensions)
Absolute mean array reduction operation, optionally along specified dimensions: out = mean(abs(x))- Parameters:
in
- Input variable (NUMERIC type)keepDims
- Whether to keep the original dimensions or produce a shrunk array with less dimensionsdimensions
- Dimensions to reduce over. If dimensions are not specified, full array reduction is performed (Size: AtLeast(min=0))- Returns:
- output Reduced array of rank (input rank - num dimensions) (NUMERIC type)
-
reduceAmean
public INDArray reduceAmean(INDArray in, int... dimensions)
Absolute mean array reduction operation, optionally along specified dimensions: out = mean(abs(x))- Parameters:
in
- Input variable (NUMERIC type)dimensions
- Dimensions to reduce over. If dimensions are not specified, full array reduction is performed (Size: AtLeast(min=0))- Returns:
- output Reduced array of rank (input rank - num dimensions) (NUMERIC type)
-
reduceAmean
public INDArray reduceAmean(INDArray in, INDArray dimensions, boolean keepDims)
Absolute mean array reduction operation, optionally along specified dimensions: out = mean(abs(x))- Parameters:
in
- Input variable (NUMERIC type)dimensions
- Dimensions to reduce along (NUMERIC type)keepDims
- Whether to keep the original dimensions or produce a shrunk array with less dimensions- Returns:
- output Reduced array of rank (input rank - num dimensions) (NUMERIC type)
-
reduceAmean
public INDArray reduceAmean(INDArray in, INDArray dimensions)
Absolute mean array reduction operation, optionally along specified dimensions: out = mean(abs(x))- Parameters:
in
- Input variable (NUMERIC type)dimensions
- Dimensions to reduce along (NUMERIC type)- Returns:
- output Reduced array of rank (input rank - num dimensions) (NUMERIC type)
-
reduceAmin
public INDArray reduceAmin(INDArray in, boolean keepDims, int... dimensions)
Absolute min array reduction operation, optionally along specified dimensions: out = min(abs(x))- Parameters:
in
- Input variable (NUMERIC type)keepDims
- Whether to keep the original dimensions or produce a shrunk array with less dimensionsdimensions
- Dimensions to reduce over. If dimensions are not specified, full array reduction is performed (Size: AtLeast(min=0))- Returns:
- output Reduced array of rank (input rank - num dimensions) (NUMERIC type)
-
reduceAmin
public INDArray reduceAmin(INDArray in, int... dimensions)
Absolute min array reduction operation, optionally along specified dimensions: out = min(abs(x))- Parameters:
in
- Input variable (NUMERIC type)dimensions
- Dimensions to reduce over. If dimensions are not specified, full array reduction is performed (Size: AtLeast(min=0))- Returns:
- output Reduced array of rank (input rank - num dimensions) (NUMERIC type)
-
reduceAmin
public INDArray reduceAmin(INDArray in, INDArray dimensions, boolean keepDims)
Absolute min array reduction operation, optionally along specified dimensions: out = min(abs(x))- Parameters:
in
- Input variable (NUMERIC type)dimensions
- Dimensions to reduce along (NUMERIC type)keepDims
- Whether to keep the original dimensions or produce a shrunk array with less dimensions- Returns:
- output Reduced array of rank (input rank - num dimensions) (NUMERIC type)
-
reduceAmin
public INDArray reduceAmin(INDArray in, INDArray dimensions)
Absolute min array reduction operation, optionally along specified dimensions: out = min(abs(x))- Parameters:
in
- Input variable (NUMERIC type)dimensions
- Dimensions to reduce along (NUMERIC type)- Returns:
- output Reduced array of rank (input rank - num dimensions) (NUMERIC type)
-
reduceMax
public INDArray reduceMax(INDArray in, boolean keepDims, int... dimensions)
The max of an array along each dimension- Parameters:
in
- Input variable (NUMERIC type)keepDims
- Whether to keep the original dimensions or produce a shrunk array with less dimensionsdimensions
- Dimensions to reduce over. If dimensions are not specified, full array reduction is performed (Size: AtLeast(min=0))- Returns:
- output Reduced array of rank (input rank - num dimensions) (NUMERIC type)
-
reduceMax
public INDArray reduceMax(INDArray in, int... dimensions)
The max of an array along each dimension- Parameters:
in
- Input variable (NUMERIC type)dimensions
- Dimensions to reduce over. If dimensions are not specified, full array reduction is performed (Size: AtLeast(min=0))- Returns:
- output Reduced array of rank (input rank - num dimensions) (NUMERIC type)
-
reduceMax
public INDArray reduceMax(INDArray in, INDArray dimensions, boolean keepDims)
The max of an array long each dimension- Parameters:
in
- Input variable (NUMERIC type)dimensions
- Dimensions to reduce along (NUMERIC type)keepDims
- Whether to keep the original dimensions or produce a shrunk array with less dimensions- Returns:
- output Reduced array of rank (input rank - num dimensions) (NUMERIC type)
-
reduceMax
public INDArray reduceMax(INDArray in, INDArray dimensions)
The max of an array long each dimension- Parameters:
in
- Input variable (NUMERIC type)dimensions
- Dimensions to reduce along (NUMERIC type)- Returns:
- output Reduced array of rank (input rank - num dimensions) (NUMERIC type)
-
reduceMin
public INDArray reduceMin(INDArray in, boolean keepDims, int... dimensions)
The minimum of an array along each dimension- Parameters:
in
- Input variable (NUMERIC type)keepDims
- Whether to keep the original dimensions or produce a shrunk array with less dimensionsdimensions
- Dimensions to reduce over. If dimensions are not specified, full array reduction is performed (Size: AtLeast(min=0))- Returns:
- output Reduced array of rank (input rank - num dimensions) (NUMERIC type)
-
reduceMin
public INDArray reduceMin(INDArray in, int... dimensions)
The minimum of an array along each dimension- Parameters:
in
- Input variable (NUMERIC type)dimensions
- Dimensions to reduce over. If dimensions are not specified, full array reduction is performed (Size: AtLeast(min=0))- Returns:
- output Reduced array of rank (input rank - num dimensions) (NUMERIC type)
-
reduceMin
public INDArray reduceMin(INDArray in, INDArray dimensions, boolean keepDims)
The minimum of an array long each dimension- Parameters:
in
- Input variable (NUMERIC type)dimensions
- Dimensions to reduce along (NUMERIC type)keepDims
- Whether to keep the original dimensions or produce a shrunk array with less dimensions- Returns:
- output Reduced array of rank (input rank - num dimensions) (NUMERIC type)
-
reduceMin
public INDArray reduceMin(INDArray in, INDArray dimensions)
The minimum of an array long each dimension- Parameters:
in
- Input variable (NUMERIC type)dimensions
- Dimensions to reduce along (NUMERIC type)- Returns:
- output Reduced array of rank (input rank - num dimensions) (NUMERIC type)
-
round
public INDArray round(INDArray x)
Element-wise round function: out = round(x).
Rounds (up or down depending on value) to the nearest integer value.- Parameters:
x
- Input variable (NUMERIC type)- Returns:
- output Output variable (NUMERIC type)
-
rsqrt
public INDArray rsqrt(INDArray x)
Element-wise reciprocal (inverse) of square root: out = 1.0 / sqrt(x)- Parameters:
x
- Input variable (NUMERIC type)- Returns:
- output Output variable (NUMERIC type)
-
rsub
public INDArray rsub(INDArray x, INDArray y)
Pairwise reverse subtraction operation, out = y - x
Note: supports broadcasting if x and y have different shapes and are broadcastable.
For example, if X has shape [1,10] and Y has shape [5,10] then op(X,Y) has output shape [5,10]
Broadcast rules are the same as NumPy: https://docs.scipy.org/doc/numpy/user/basics.broadcasting.html- Parameters:
x
- Input variable (NUMERIC type)y
- Input variable (NUMERIC type)- Returns:
- output Output variable (NUMERIC type)
-
rsub
public INDArray rsub(INDArray x, double value)
Scalar reverse subtraction operation, out = scalar - in- Parameters:
x
- Input variable (NUMERIC type)value
- Scalar value for op- Returns:
- output Output variable (NUMERIC type)
-
setDiag
public INDArray setDiag(INDArray in, INDArray diag)
Set the diagonal value to the specified values
If input is
[ a, b, c]
[ d, e, f]
[ g, h, i]
and diag = [ 1, 2, 3] then output is
[ 1, b, c]
[ d, 2, f]
[ g, h, 3]- Parameters:
in
- Input variable (NUMERIC type)diag
- Diagonal (NUMERIC type)- Returns:
- output Output variable (NUMERIC type)
-
shannonEntropy
public INDArray shannonEntropy(INDArray in, boolean keepDims, int... dimensions)
Shannon Entropy reduction: -sum(x * log2(x))- Parameters:
in
- Input variable (NUMERIC type)keepDims
- Whether to keep the original dimensions or produce a shrunk array with less dimensionsdimensions
- Dimensions to reduce over. If dimensions are not specified, full array reduction is performed (Size: AtLeast(min=0))- Returns:
- output Reduced array of rank (input rank - num dimensions) (NUMERIC type)
-
shannonEntropy
public INDArray shannonEntropy(INDArray in, int... dimensions)
Shannon Entropy reduction: -sum(x * log2(x))- Parameters:
in
- Input variable (NUMERIC type)dimensions
- Dimensions to reduce over. If dimensions are not specified, full array reduction is performed (Size: AtLeast(min=0))- Returns:
- output Reduced array of rank (input rank - num dimensions) (NUMERIC type)
-
shannonEntropy
public INDArray shannonEntropy(INDArray in, INDArray dimensions, boolean keepDims)
Shannon Entropy reduction: -sum(x * log2(x))- Parameters:
in
- Input variable (NUMERIC type)dimensions
- Dimensions to reduce along (NUMERIC type)keepDims
- Whether to keep the original dimensions or produce a shrunk array with less dimensions- Returns:
- output Reduced array of rank (input rank - num dimensions) (NUMERIC type)
-
shannonEntropy
public INDArray shannonEntropy(INDArray in, INDArray dimensions)
Shannon Entropy reduction: -sum(x * log2(x))- Parameters:
in
- Input variable (NUMERIC type)dimensions
- Dimensions to reduce along (NUMERIC type)- Returns:
- output Reduced array of rank (input rank - num dimensions) (NUMERIC type)
-
sign
public INDArray sign(INDArray x)
Element-wise sign (signum) function:
out = -1 if in < 0
out = 0 if in = 0
out = 1 if in > 0- Parameters:
x
- Input variable (NUMERIC type)- Returns:
- output Output variable (NUMERIC type)
-
sin
public INDArray sin(INDArray x)
Elementwise sine operation: out = sin(x)- Parameters:
x
- Input variable (NUMERIC type)- Returns:
- output Output variable (NUMERIC type)
-
sinh
public INDArray sinh(INDArray x)
Elementwise sinh (hyperbolic sine) operation: out = sinh(x)- Parameters:
x
- Input variable (NUMERIC type)- Returns:
- output Output variable (NUMERIC type)
-
sqrt
public INDArray sqrt(INDArray x)
Element-wise square root function: out = sqrt(x)- Parameters:
x
- Input variable (NUMERIC type)- Returns:
- output Output variable (NUMERIC type)
-
square
public INDArray square(INDArray x)
Element-wise square function: out = x^2- Parameters:
x
- Input variable (NUMERIC type)- Returns:
- output Output variable (NUMERIC type)
-
squaredDifference
public INDArray squaredDifference(INDArray x, INDArray y)
Pairwise squared difference operation.
Note: supports broadcasting if x and y have different shapes and are broadcastable.
For example, if X has shape [1,10] and Y has shape [5,10] then op(X,Y) has output shape [5,10]
Broadcast rules are the same as NumPy: https://docs.scipy.org/doc/numpy/user/basics.broadcasting.html- Parameters:
x
- Input variable (NUMERIC type)y
- Input variable (NUMERIC type)- Returns:
- output Output variable (NUMERIC type)
-
squaredNorm
public INDArray squaredNorm(INDArray in, boolean keepDims, int... dimensions)
Sum of squared differences.- Parameters:
in
- Input variable (NUMERIC type)keepDims
- Whether to keep the original dimensions or produce a shrunk array with less dimensionsdimensions
- Dimensions to reduce over. If dimensions are not specified, full array reduction is performed (Size: AtLeast(min=0))- Returns:
- output Reduced array of rank (input rank - num dimensions) (NUMERIC type)
-
squaredNorm
public INDArray squaredNorm(INDArray in, int... dimensions)
Sum of squared differences.- Parameters:
in
- Input variable (NUMERIC type)dimensions
- Dimensions to reduce over. If dimensions are not specified, full array reduction is performed (Size: AtLeast(min=0))- Returns:
- output Reduced array of rank (input rank - num dimensions) (NUMERIC type)
-
squaredNorm
public INDArray squaredNorm(INDArray in, INDArray dimensions, boolean keepDims)
Sum of squared differences.- Parameters:
in
- Input variable (NUMERIC type)dimensions
- Dimensions to reduce along (NUMERIC type)keepDims
- Whether to keep the original dimensions or produce a shrunk array with less dimensions- Returns:
- output Reduced array of rank (input rank - num dimensions) (NUMERIC type)
-
squaredNorm
public INDArray squaredNorm(INDArray in, INDArray dimensions)
Sum of squared differences.- Parameters:
in
- Input variable (NUMERIC type)dimensions
- Dimensions to reduce along (NUMERIC type)- Returns:
- output Reduced array of rank (input rank - num dimensions) (NUMERIC type)
-
standardize
public INDArray standardize(INDArray x, int... dimensions)
Standardize input variable along given axis
out = (x - mean) / stdev
with mean and stdev being calculated along the given dimension.
For example: given x as a mini batch of the shape [numExamples, exampleLength]:
- use dimension 1 too use the statistics (mean, stdev) for each example
- use dimension 0 if you want to use the statistics for each column across all examples
- use dimensions 0,1 if you want to use the statistics across all columns and examples
- Parameters:
x
- Input variable (NUMERIC type)dimensions
- (Size: AtLeast(min=1))- Returns:
- output Output variable (NUMERIC type)
-
step
public INDArray step(INDArray x, double value)
Elementwise step function:
out(x) = 1 if x >= cutoff
out(x) = 0 otherwise- Parameters:
x
- Input variable (NUMERIC type)value
- Scalar value for op- Returns:
- output Output variable (NUMERIC type)
-
sub
public INDArray sub(INDArray x, INDArray y)
Pairwise subtraction operation, out = x - y
Note: supports broadcasting if x and y have different shapes and are broadcastable.
For example, if X has shape [1,10] and Y has shape [5,10] then op(X,Y) has output shape [5,10]
Broadcast rules are the same as NumPy: https://docs.scipy.org/doc/numpy/user/basics.broadcasting.html- Parameters:
x
- Input variable (NUMERIC type)y
- Input variable (NUMERIC type)- Returns:
- output Output variable (NUMERIC type)
-
sub
public INDArray sub(INDArray x, double value)
Scalar subtraction operation, out = in - scalar- Parameters:
x
- Input variable (NUMERIC type)value
- Scalar value for op- Returns:
- output Output variable (NUMERIC type)
-
sum
public INDArray sum(INDArray in, boolean keepDims, int... dimensions)
Sum of an array, optionally along specified dimensions: out = sum(x))- Parameters:
in
- Input variable (NUMERIC type)keepDims
- Whether to keep the original dimensions or produce a shrunk array with less dimensionsdimensions
- Dimensions to reduce over. If dimensions are not specified, full array reduction is performed (Size: AtLeast(min=0))- Returns:
- output Reduced array of rank (input rank - num dimensions) (NUMERIC type)
-
sum
public INDArray sum(INDArray in, int... dimensions)
Sum of an array, optionally along specified dimensions: out = sum(x))- Parameters:
in
- Input variable (NUMERIC type)dimensions
- Dimensions to reduce over. If dimensions are not specified, full array reduction is performed (Size: AtLeast(min=0))- Returns:
- output Reduced array of rank (input rank - num dimensions) (NUMERIC type)
-
sum
public INDArray sum(INDArray in, INDArray dimensions, boolean keepDims)
Sum of an array, optionally along specified dimensions: out = sum(x))- Parameters:
in
- Input variable (NUMERIC type)dimensions
- Dimensions to reduce along (NUMERIC type)keepDims
- Whether to keep the original dimensions or produce a shrunk array with less dimensions- Returns:
- output Reduced array of rank (input rank - num dimensions) (NUMERIC type)
-
sum
public INDArray sum(INDArray in, INDArray dimensions)
Sum of an array, optionally along specified dimensions: out = sum(x))- Parameters:
in
- Input variable (NUMERIC type)dimensions
- Dimensions to reduce along (NUMERIC type)- Returns:
- output Reduced array of rank (input rank - num dimensions) (NUMERIC type)
-
tan
public INDArray tan(INDArray x)
Elementwise tangent operation: out = tan(x)- Parameters:
x
- Input variable (NUMERIC type)- Returns:
- output Output variable (NUMERIC type)
-
tanh
public INDArray tanh(INDArray x)
Elementwise tanh (hyperbolic tangent) operation: out = tanh(x)- Parameters:
x
- Input variable (NUMERIC type)- Returns:
- output Output variable (NUMERIC type)
-
trace
public INDArray trace(INDArray in)
Matrix trace operation
For rank 2 matrices, the output is a scalar with the trace - i.e., sum of the main diagonal.
For higher rank inputs, output[a,b,c] = trace(in[a,b,c,:,:])- Parameters:
in
- Input variable (NUMERIC type)- Returns:
- output Trace (NUMERIC type)
-
xor
public INDArray xor(INDArray x, INDArray y)
Boolean XOR (exclusive OR) operation: elementwise (x != 0) XOR (y != 0)
If x and y arrays have equal shape, the output shape is the same as these inputs.
Note: supports broadcasting if x and y have different shapes and are broadcastable.
Returns an array with values 1 where condition is satisfied, or value 0 otherwise.- Parameters:
x
- Input 1 (BOOL type)y
- Input 2 (BOOL type)- Returns:
- output INDArray with values 0 and 1 based on where the condition is satisfied (BOOL type)
-
-