I have a question about convolution kernel.
Is there a way to restraint the kernel(4,4) and it’s weight between 0-1 and their sum is 1, which ensures that this condition can always be met during training.
I have a question about convolution kernel.
Is there a way to restraint the kernel(4,4) and it’s weight between 0-1 and their sum is 1, which ensures that this condition can always be met during training.