How does tiling work in convolutional neural networks? I get the basics of how convolution layers work, but all the papers I read seem to talk about tiling as if it’s sampled differently.