Flexible output sizes for jax-unirep

PR: https://github.com/ElArkk/jax-unirep/pull/80

Goals:

• We want flexible number of dimensions for outputs. This should help with training models on large amounts of data.
• A slightly more functional API for the fit function.
• Update docs.

This is an overview page of my notes on differential computing and JAX.

Contents:

1. What is differential computing
2. What automatic differentiation systems exist
3. Where differential computing gets used
4. Differential computing with JAX
5. What makes JAX special
6. Composable program transforms
7. Infinitely Wide Neural Networks

Stuff I've built/done with JAX:

1. jax-unirep
2. Workshop on Differential Learning
3. Estimating a multivariate Gaussian's parameters by gradient descent

This is for the jax-unirep paper.

https://peerj.com/about/policies-and-procedures/#discipline-standards

PeerJ welcomes articles describing bioinformatics software tools. These articles should present new software tools (or significant new functionality in existing software) of particular interest to the bioinformatics and/or computational biology communities. The described tools should provide new computational or analytical functionality for researchers.

The functionality of the software should, where appropriate, be validated using real-world biological data and/or compared to existing tools. If available as a package (e.g. Python, R, Matlab, Octave), it should be accompanied by a minimal script that downloads the data (if not included in the package), loads it, and performs the analysis to reproduce the results (tables, plots, visualizations etc.) in the manuscript. Documentation & comments must be clear and sufficient to allow a typical user to perform the analysis described in the article. Ideally, the tools should be usable in a workflow that encourages reproducible research practices.

The software must be released under an open source license (e.g. MIT, GPL) and be widely available (i.e. hosted in a public repository such as Github or Bitbucket, or an institutional repository). Use of a version control system (e.g. Git, Subversion etc.) is strongly encouraged, as is adherence to language-specific packaging practices where appropriate. We also recommend the inclusion of appropriate unit tests. The software must be free to noncommercial users, and must be accessible without requiring any personally identifiable information. Software and validation data sets must comply with PeerJ Data and Materials Sharing policies. Reliance on a proprietary software such as Matlab does not preclude the publication but in general, a fully open source method is to be preferred.

The guidelines are good. Scope is right. I say we go for it.