Consider the following excerpt taken from the chapter named Using convolutions to generalize from the textbook titled Deep Learning with PyTorch by Eli Stevens et al.
Downsampling could in principle occur in different ways. Scaling an image by half is the equivalent of taking four neighboring pixels as input and producing one pixel as output. How we compute the value of the output based on the values of the input is up to us. We could
- Average the four pixels. This average pooling was a common approach early on but has fallen out of favor somewhat.
- Take the maximum of the four pixels. This approach, called max pooling, is currently the most commonly used approach, but it has a downside of discarding the other three-quarters of the data.
- Perform a strided convolution, where only every $N$-th pixel is calculated. A $3 \times 4$ convolution with stride 2 still incorporates input from all pixels from the previous layer. The literature shows promise for this approach, but it has not yet supplanted max pooling.
The paragraph is mentioning that the research community is biased towards max-pooling than average pooling. Is there any rational basis for such bias?