Experiment to attempt to use backpropagation to see how a convolution layer in a CNN actually converges to a kernel filter. In this case trying to converge to a sobel filter for edge detection. Experiment findings are in sobel juypter notebook

References: https://pavisj.medium.com/convolutions-and-backpropagations-46026a8f5d2c. Huge credit for understanding how backpropogation works in a convolution layer. https://www.tomasbeuzen.com/deep-learning-with-pytorch/chapters/chapter5_cnns-pt1.html. Great visualizations