Machine Learning Exploration

Projects to explore and develop ML intuition.

Transfer Learning with Convolutional Neural Nets

Use a VGG architecture with pre-trained weights to apply the style of one image to another with a simple transformation. This takes some tuning to get good results.

Decoding Ciphers with Markov Chain Monte Carlo (MCMC)

Contains the entirety of Moby Dick, chapter by chapter. Using a substitution cipher, encode chapters the book into a scrambled language, then efficiently recover the original text using MCMC.

K-Nearest Neighbors from Scratch

A walkthrough of one of the most simple classification algorithms, with intuition how how it works geometrically.

Vanilla Neural Networks and MNIST with Tensorboard

Train simple neural networks in Keras with a Tensorflow backend. Interpret the model's evolution at train-time through a local TensorBoard instance.