Mixture Density Networks with Julia
This post is also available as a Jupyter notebook.
Related posts:
JavaScript implementation. TensorFlow implementation. PyTorch implementation This post is a result of me trying to understand how to do deep learning in Julia using the excellent Flux package as well as getting a better understanding of conditional density estimation using a simple but effective technique—Mixture Density Networks (Bishop, 1994). This post follows very closely the PyTorch implementation (including paraphrasing some statements) listed above, which itself was adapted from the original TensorFlow and JavaScript implementations.
[Read More]