Probabilistic Models with Deep Neural Networks

This repository refers to the following manuscript:

• Title: Probabilistic Models with Deep Neural Networks
• Authors: Andrés R. Masegosa, Rafael Cabañas, Helge Langseth, Thomas D. Nielsen, Antonio Salmerón
• PDF: https://arxiv.org/pdf/1908.03442.pdf

Here we provide the code for running the examples in the aforementioned manuscript using TensorFlow Probability. These are provided in Jupyter notebook format and can be executed in cloud services:

• Example 1 - PCA
• Example 6 - Non-linear PCA
• Example 10 - VAE

Another two well known probabilistic models including artificial neural newtorks, not detailed in the manuscript, can be found here:

• Bayesian Neural Networks
• Mixture Density Networks

The same notebooks are also coded using InferPy, which is a high-level API for probabilistic modeling with deep neural networks. InferPy has a strong focus on ease of use.

• Example 1 - PCA
• Example 6 - Non-linear PCA
• Example 10 - VAE
• Bayesian Neural Networks
• Mixture Density Networks

A comparison of the evolution of the ELBO (i.e., objective function) is shown in the following notebook:

• ELBO Evolution

Citation:

@article{masegosa2019probabilistic,
  author = {Masegosa, A.R. and Caba\~{n}as, R. and Langseth, H. and Nielsen, T.D. and Salmer\'{o}n, A. },
    title = {Probabilistic Models with Deep Neural Networks},
  journal = {arXiv preprint arXiv:1908.03442},
  year = {2019}
}