What changes in settings would be necessary to train the network using a phi matrix (block size) bigger than 33x33, say 128x128? I successfully trained the network with block size of 66x66, but bigger values are extremely difficult to train. Any thoughts?
What changes in settings would be necessary to train the network using a phi matrix (block size) bigger than 33x33, say 128x128? I successfully trained the network with block size of 66x66, but bigger values are extremely difficult to train. Any thoughts?