This repository contains various demonstrations of the Flux machine learning library. Any of these may freely be used as a starting point for your own models.
- housing implements the most basic model possible (a linear regression) on the UCI housing data set. It's bare-bones and illustrates how to build a model from scratch.
- mnist classifies digits from the MNIST data set, using a simple multi-layer perceptron.
- char-rnn implements a character-level language model. It comes with a Shakespeare dataset but can work with any text.
- phonemes implements a sequence to sequence model with attention, using the CMU pronouncing dictionary to predict the pronunciations of unknown words.
Note that these models are best run line-by-line, either in the REPL or Juno.