I spent whole day trying to figure out how convn function works in 'valid' mode for order 3 tensor. I have an input i of size 3*3*2 and a filter w of size 2*2*2 but the output size after convolution is 2*2 as shown below. I just need to understand how the dot product of convolution was performed on both channels of the input 1. please if you can demonstrate how the first entry of the output was calculated by convn function. note that i am rotating w 180 degrees so as to perform cross correlation instead of convolution just for ease of hand calculations check up. many thanks in advance.