deepnetts.util

Class Tensor

java.lang.Object

deepnetts.util.Tensor

All Implemented Interfaces:

Serializable

public class Tensor
extends Object
implements Serializable

This class represents a wrapper for multidimensional array. It a basic data structure for storing inputs, outputs and internal parameters(weights) of the deep learning model. It supports up to 4 dimensions - it can be 1D, 2D, 3D or 4D.

See Also:

Serialized Form

Constructor Summary

Constructors

Constructor and Description
Tensor(float... values) Creates a single row tensor with specified values.
Tensor(float[][] vals) Creates a 2D tensor / matrix with specified values.
Tensor(float[][][] vals) Creates a 3D tensor from specified 3D array
Tensor(float[][][][] vals)
Tensor(int cols) Creates an empty single row tensor with specified number of columns.
Tensor(int cols, float val)
Tensor(int rows, int cols) Creates a tensor with specified number of rows and columns.
Tensor(int rows, int cols, float[] values)
Tensor(int rows, int cols, int depth) Creates a 3D tensor with specified number of rows, cols and depth.
Tensor(int rows, int cols, int depth, float[] values)
Tensor(int rows, int cols, int depth, int fourthDim)
Tensor(int rows, int cols, int depth, int fourthDim, float[] values)
Tensor(Shape shape, float... values)
Tensor(Tensor t) Public deep copy / clone constructor.

Method Summary

Modifier and Type	Method and Description
void	add(float value, int idx)
void	add(float value, int row, int col) Adds specified value to matrix value at position x, y
void	add(float value, int row, int col, int z)
void	add(float value, int row, int col, int z, int fourth)
void	add(Tensor t) Adds specified tensor t to this tensor.
void	apply(Function<Float,Float> f)
Tensor	clone()
Tensor	copy()
void	copyFrom(float[] src)
static Tensor	create(int rows, int cols, float[] values) Factory method for creating tensor instance,
static Tensor	create(int rows, int cols, int depth, float[] values)
static Tensor	create(int rows, int cols, int depth, int fourthDim, float[] values)
void	div(float value) Divide all values in this tensor with specified value.
void	div(float[] divisors)
void	div(Tensor t)
boolean	equals(Object obj)
boolean	equals(Tensor t2, float delta)
void	fill(float value) Fills the entire tensor with specified value.
static void	fill(float[] array, float val)
float	get(int idx) Gets value at specified index position.
float	get(int row, int col) Returns matrix value at row, col
float	get(int row, int col, int z) Returns value at specified index [row, col, z].
float	get(int row, int col, int z, int fourth)
int	getCols()
int	getDepth()
int	getFourthDim()
int	getRank() Rank corresponds to number of dimensions in tensor.
int	getRows()
float[]	getValues()
float	getWithStride(int[] idxs)
int	hashCode()
float	mean()
void	multiply(float m)
void	multiply(float mul, int idx)
void	multiplyElementWise(Tensor tensor2)
static Tensor	of(float[] values)
void	randomize() Randomize all values in tensor
void	replace(float toReplace, float replaceWith)
void	set(float val, int idx) Sets value at specified index position.
void	set(float val, int row, int col) Set value at specified [row, col] position.
void	set(float val, int row, int col, int z)
void	set(float val, int row, int col, int z, int fourth)
void	setValues(float... values)
void	setValuesFromString(String values) Sets tensor values from CSV string.
int	size() Total number of values in tensor.
void	sqrt()
float	std(float mean)
void	sub(float val)
void	sub(float val, int row, int col)
void	sub(float val, int row, int col, int z)
void	sub(float val, int row, int col, int z, int fourth)
void	sub(Tensor t) Subtracts specified tensor t from this tensor.
static void	sub(Tensor t1, Tensor t2) Subtracts tensor t2 from t1.
float	sumAbs() Returns sum of abs values of this tensor - L1 norm
float	sumSqr() Returns sum of squared values of this tensor - L2 norm
String	toString()
static String	valuesAsString(Tensor[] tensors)

Methods inherited from class java.lang.Object

getClass, notify, notifyAll, wait, wait, wait

Constructor Detail

Tensor

public Tensor(float... values)

Creates a single row tensor with specified values.

Parameters:

values - values of column tensor

Tensor

public Tensor(Shape shape,
              float... values)

Tensor

public Tensor(float[][] vals)

Creates a 2D tensor / matrix with specified values.

Parameters:

vals - values to put into matrix

Tensor

public Tensor(float[][][] vals)

Creates a 3D tensor from specified 3D array

Parameters:

vals - 2D array of tensor values

Tensor

public Tensor(float[][][][] vals)

Tensor

public Tensor(int cols)

Creates an empty single row tensor with specified number of columns.

Parameters:

cols - number of columns

Tensor

public Tensor(int cols,
              float val)

Tensor

public Tensor(int rows,
              int cols)

Creates a tensor with specified number of rows and columns.

Parameters:

rows - number of rows

cols - number of columns

Tensor

public Tensor(int rows,
              int cols,
              float[] values)

Tensor

public Tensor(int rows,
              int cols,
              int depth)

Creates a 3D tensor with specified number of rows, cols and depth.

Parameters:

rows - number of rows

cols - number of columns

depth - tensor depth

Tensor

public Tensor(int rows,
              int cols,
              int depth,
              int fourthDim)

Tensor

public Tensor(int rows,
              int cols,
              int depth,
              int fourthDim,
              float[] values)

Tensor

public Tensor(int rows,
              int cols,
              int depth,
              float[] values)

Tensor

public Tensor(Tensor t)

Public deep copy / clone constructor.

Parameters:

t -

Method Detail

clone

public Tensor clone()

Overrides:

clone in class Object

get

public final float get(int idx)

Gets value at specified index position.

Parameters:

idx -

Returns:

set

public final void set(float val,
                      int idx)

Sets value at specified index position.

Parameters:

idx -

val -

get

public final float get(int row,
                       int col)

Returns matrix value at row, col

Parameters:

col -

row -

Returns:

value at [row, col]

set

public final void set(float val,
                      int row,
                      int col)

Set value at specified [row, col] position.

Parameters:

val - value to set

row - tensor's row index

col - tensor's col index

get

public final float get(int row,
                       int col,
                       int z)

Returns value at specified index [row, col, z].

Parameters:

col -

row -

z -

Returns:

set

public final void set(float val,
                      int row,
                      int col,
                      int z)

get

public final float get(int row,
                       int col,
                       int z,
                       int fourth)

set

public final void set(float val,
                      int row,
                      int col,
                      int z,
                      int fourth)

getWithStride

public final float getWithStride(int[] idxs)

getValues

public final float[] getValues()

setValues

public final void setValues(float... values)

copyFrom

public final void copyFrom(float[] src)

getCols

public final int getCols()

getRows

public final int getRows()

getDepth

public final int getDepth()

getFourthDim

public final int getFourthDim()

getRank

public final int getRank()

Rank corresponds to number of dimensions in tensor.

Returns:

size

public final int size()

Total number of values in tensor.

Returns:

toString

public String toString()

Overrides:

toString in class Object

add

public final void add(float value,
                      int idx)

add

public final void add(float value,
                      int row,
                      int col)

Adds specified value to matrix value at position x, y

Parameters:

col -

row -

value -

add

public final void add(float value,
                      int row,
                      int col,
                      int z)

add

public final void add(float value,
                      int row,
                      int col,
                      int z,
                      int fourth)

add

public final void add(Tensor t)

Adds specified tensor t to this tensor.

Parameters:

t - tensor to add

sub

public final void sub(float val,
                      int row,
                      int col)

sub

public final void sub(float val,
                      int row,
                      int col,
                      int z)

sub

public final void sub(float val,
                      int row,
                      int col,
                      int z,
                      int fourth)

sub

public final void sub(Tensor t)

Subtracts specified tensor t from this tensor.

Parameters:

t - tensor to subtract

sub

public final void sub(float val)

sub

public static final void sub(Tensor t1,
                             Tensor t2)

Subtracts tensor t2 from t1. The result is t1.

Parameters:

t1 -

t2 -

div

public final void div(float value)

Divide all values in this tensor with specified value.

Parameters:

value -

div

public final void div(float[] divisors)

fill

public final void fill(float value)

Fills the entire tensor with specified value.

Parameters:

value - value used to fill tensor

fill

public static final void fill(float[] array,
                              float val)

div

public final void div(Tensor t)

copy

public Tensor copy()

apply

public void apply(Function<Float,Float> f)

equals

public boolean equals(Object obj)

Overrides:

equals in class Object

hashCode

public int hashCode()

Overrides:

hashCode in class Object

equals

public boolean equals(Tensor t2,
                      float delta)

valuesAsString

public static String valuesAsString(Tensor[] tensors)

setValuesFromString

public void setValuesFromString(String values)

Sets tensor values from CSV string.

Parameters:

values - string with comma separated tensor values

Throws:

exception - if number of values in string does not match number of values in tensor

create

public static Tensor create(int rows,
                            int cols,
                            float[] values)

Factory method for creating tensor instance,

Parameters:

rows -

cols -

values -

Returns:

create

public static Tensor create(int rows,
                            int cols,
                            int depth,
                            float[] values)

create

public static Tensor create(int rows,
                            int cols,
                            int depth,
                            int fourthDim,
                            float[] values)

of

public static Tensor of(float[] values)

sumAbs

public float sumAbs()

Returns sum of abs values of this tensor - L1 norm

Returns:

L1 norm

sumSqr

public float sumSqr()

Returns sum of squared values of this tensor - L2 norm

Returns:

L2 norm

randomize

public void randomize()

Randomize all values in tensor

multiplyElementWise

public void multiplyElementWise(Tensor tensor2)

multiply

public void multiply(float m)

multiply

public void multiply(float mul,
                     int idx)

sqrt

public void sqrt()

mean

public float mean()

std

public float std(float mean)

replace

public void replace(float toReplace,
                    float replaceWith)