Follow-up Post about MDCT Synth built w/ Autoencoder

[JohnVinyard]

Write a follow-up to this post: https://johnvinyard.github.io/zounds/synthesis/2016/08/13/building-a-simple-audio-encoder.html

Instead of K-means clustering, use keras to build an autoencoder, as described here: https://blog.keras.io/building-autoencoders-in-keras.html