This is a debug view that shows how a generative model works with the MNIST dataset of handwritten digits. The model is a Semi-Supervised Variational Auto-Encoder (SSVAE). There is an Encoder and a Decoder, as well as a Classifier. I step through the iterations, and we can see the error and accuracy improve, and the generated digits gradually transform from white noise to clear numbers from 0 through 9. This helped me debug an SSVAE model for a class project, and really understand how it works.
