Implement max/mean pooling(with stride) with numpy

Question

I wanted to know how to implement a simple max/mean pooling with numpy. I was reading Max and mean pooling with numpy, but unfortunately it assumed the stride was the same as the kernel size. Is there a numpythonic way to do this? Also it would be nice if this were to work for any dimension, but of course not neccesary.

Answer 1

Here's a pure numpy implementation using stride_tricks:

import numpy as np
from numpy.lib.stride_tricks import as_strided


def pool2d(A, kernel_size, stride, padding=0, pool_mode='max'):
   '''
    2D Pooling

    Parameters:
        A: input 2D array
        kernel_size: int, the size of the window over which we take pool
        stride: int, the stride of the window
        padding: int, implicit zero paddings on both sides of the input
        pool_mode: string, 'max' or 'avg'
    '''
    # Padding
    A = np.pad(A, padding, mode='constant')

    # Window view of A
    output_shape = ((A.shape[0] - kernel_size) // stride + 1,
                    (A.shape[1] - kernel_size) // stride + 1)
    
    shape_w = (output_shape[0], output_shape[1], kernel_size, kernel_size)
    strides_w = (stride*A.strides[0], stride*A.strides[1], A.strides[0], A.strides[1])
    
    A_w = as_strided(A, shape_w, strides_w)

    # Return the result of pooling
    if pool_mode == 'max':
        return A_w.max(axis=(2, 3))
    elif pool_mode == 'avg':
        return A_w.mean(axis=(2, 3))

Example:

>>> A = np.array([[1, 1, 2, 4],
                  [5, 6, 7, 8],
                  [3, 2, 1, 0],
                  [1, 2, 3, 4]])

>>> pool2d(A, kernel_size=2, stride=2, padding=0, pool_mode='max')

array([[6, 8],
       [3, 4]])

https://cs231n.github.io/convolutional-networks/