machine-learning - calculate number of unique weights in a convolution layer

问题描述

If there is an input of 6x6 pixels and 3 channels and the padding and stride are both equal to 1 and the kernel size is 4x4. The output should only have 1 channel, how do you compute the number of unique weights in the layer?

I know that the formula to compute the output is subtracting the kernel size from the pixels and adding 2 times the padding, then dividing this number by the strides and adding 1. However I cannot seem to find a way to calculate the number of weights in a layer, could someone help me with this?

标签： machine-learningconv-neural-network

machine-learning - calculate number of unique weights in a convolution layer

问题描述

解决方案

推荐阅读