Convolution in Two Dimensions
Dive into a comprehensive 2.5-hour tutorial on two-dimensional convolution and convolutional neural networks. Learn the fundamentals of convolution operations, explore the Softmax neural network layer, and understand batch normalization. Build a 2D convolutional neural network from scratch, starting with the basics and progressing through a step-by-step implementation using the MNIST digits dataset. Gain insights into model architecture, pre-trained model results, and analyze examples of successes and failures. Discover the advantages of using Cottonwood framework, set up training code, add and connect layers, and implement training and testing loops. Explore loss history visualization, text summaries, and learn how to collect and render examples. Conclude with a Cottonwood code tour and a handy cheatsheet for quick reference.