ai.djl.nn.pooling

Class Pool

java.lang.Object

ai.djl.nn.pooling.Pool

public final class Pool
extends java.lang.Object

Utility class that provides Block and methods for different pooling functions.

Method Summary

Modifier and Type	Method and Description
static NDArray	avgPool(NDArray data, Shape kernel, Shape stride, Shape pad) Performs Avg Pooling on the input.
static NDArray	avgPool(NDArray data, Shape kernel, Shape stride, Shape pad, PoolingConvention poolingConvention, boolean countIncludePad) Performs Avg Pooling on the input.
static NDArray	globalAvgPool(NDArray data) Performs Global Avg Pooling on the input.
static NDArray	globalLpPool(NDArray data, int pValue) Performs Global LP Pooling on the input.
static NDArray	globalMaxPool(NDArray data) Performs Global Max Pooling on the input.
static NDArray	globalSumPool(NDArray data) Performs Global Sum Pooling on the input.
static NDArray	lpPool(NDArray data, Shape kernel, Shape stride, Shape pad, int pValue) Performs LP Pooling on the input.
static NDArray	lpPool(NDArray data, Shape kernel, Shape stride, Shape pad, PoolingConvention poolingConvention, int pValue) Performs LP Pooling on the input.
static NDArray	maxPool(NDArray data, Shape kernel, Shape stride, Shape pad) Performs Max Pooling on the input.
static NDArray	maxPool(NDArray data, Shape kernel, Shape stride, Shape pad, PoolingConvention poolingConvention) Performs Max Pooling on the input.
static NDArray	sumPool(NDArray data, Shape kernel, Shape stride, Shape pad) Performs Sum Pooling on the input.
static NDArray	sumPool(NDArray data, Shape kernel, Shape stride, Shape pad, PoolingConvention poolingConvention) Performs Sum Pooling on the input.

Methods inherited from class java.lang.Object

clone, equals, finalize, getClass, hashCode, notify, notifyAll, toString, wait, wait, wait

Method Detail

maxPool

public static NDArray maxPool(NDArray data,
                              Shape kernel,
                              Shape stride,
                              Shape pad,
                              PoolingConvention poolingConvention)

Performs Max Pooling on the input.

Parameters:

data - the NDArray on which max pooling is performed

kernel - the shape of the kernel to be used

stride - the stride to be used for each dimension

pad - the padding to be set in each dimension

poolingConvention - the pooling convention to be used

Returns:

the NDArray after applying max pooling

maxPool

public static NDArray maxPool(NDArray data,
                              Shape kernel,
                              Shape stride,
                              Shape pad)

Performs Max Pooling on the input.

Parameters:

data - the NDArray on which max pooling is performed

kernel - the shape of the kernel to be used

stride - the stride to be used for each dimension

pad - the padding to be set in each dimension

Returns:

the NDArray after applying max pooling

globalMaxPool

public static NDArray globalMaxPool(NDArray data)

Performs Global Max Pooling on the input.

Parameters:

data - the NDArray on which max pooling is performed

Returns:

the NDArray after applying global max pooling

sumPool

public static NDArray sumPool(NDArray data,
                              Shape kernel,
                              Shape stride,
                              Shape pad,
                              PoolingConvention poolingConvention)

Performs Sum Pooling on the input.

Parameters:

data - the NDArray on which sum pooling is performed

kernel - the shape of the kernel to be used

stride - the stride to be used for each dimension

pad - the padding to be set in each dimension

poolingConvention - the pooling convention to be used

Returns:

the NDArray after applying sum pooling

sumPool

public static NDArray sumPool(NDArray data,
                              Shape kernel,
                              Shape stride,
                              Shape pad)

Performs Sum Pooling on the input.

Parameters:

data - the NDArray on which sum pooling is performed

kernel - the shape of the kernel to be used

stride - the stride to be used for each dimension

pad - the padding to be set in each dimension

Returns:

the NDArray after applying sum pooling

globalSumPool

public static NDArray globalSumPool(NDArray data)

Performs Global Sum Pooling on the input.

Parameters:

data - the NDArray on which sum pooling is performed

Returns:

the NDArray after applying global sum pooling

avgPool

public static NDArray avgPool(NDArray data,
                              Shape kernel,
                              Shape stride,
                              Shape pad,
                              PoolingConvention poolingConvention,
                              boolean countIncludePad)

Performs Avg Pooling on the input.

Parameters:

data - the NDArray on which average pooling is performed

kernel - the shape of the kernel to be used

stride - the stride to be used for each dimension

pad - the padding to be set in each dimension

poolingConvention - the pooling convention to be used

countIncludePad - whether to include padding for calculations

Returns:

the NDArray after applying avg pooling

avgPool

public static NDArray avgPool(NDArray data,
                              Shape kernel,
                              Shape stride,
                              Shape pad)

Performs Avg Pooling on the input.

Parameters:

data - the NDArray on which average pooling is performed

kernel - the shape of the kernel to be used

stride - the stride to be used for each dimension

pad - the padding to be set in each dimension

Returns:

the NDArray after applying avg pooling

globalAvgPool

public static NDArray globalAvgPool(NDArray data)

Performs Global Avg Pooling on the input.

Parameters:

data - the NDArray on which average pooling is performed

Returns:

the NDArray after applying global avg pooling

lpPool

public static NDArray lpPool(NDArray data,
                             Shape kernel,
                             Shape stride,
                             Shape pad,
                             PoolingConvention poolingConvention,
                             int pValue)

Performs LP Pooling on the input.

Parameters:

data - the NDArray on which LP pooling is performed

kernel - the shape of the kernel to be used

stride - the stride to be used for each dimension

pad - the padding to be set in each dimension

poolingConvention - the pooling convention to be used

pValue - the power of the pooling

Returns:

the NDArray after applying lp pooling

lpPool

public static NDArray lpPool(NDArray data,
                             Shape kernel,
                             Shape stride,
                             Shape pad,
                             int pValue)

Performs LP Pooling on the input.

Parameters:

data - the NDArray on which LP pooling is performed

kernel - the shape of the kernel to be used

stride - the stride to be used for each dimension

pad - the padding to be set in each dimension

pValue - the power of the pooling

Returns:

the NDArray after applying lp pooling

globalLpPool

public static NDArray globalLpPool(NDArray data,
                                   int pValue)

Performs Global LP Pooling on the input.

Parameters:

data - the NDArray on which LP pooling is performed

pValue - the power of the pooling

Returns:

the NDArray after applying global lp pooling