search

ubcs3 / 2017-Winter

UBC Scientific Software Seminar: Neural Networks and Deep Learning in Python
7 stars 3 forks source link

Specific examples of scientific-computing related OSS contribution #3

Open asberk opened 7 years ago

asberk commented 7 years ago

Here are two [very particular] examples of contributing a tool to a major oss project, scipy. They are both examples of tools that have a scientific computing flavour — tools that have some deeper mathematical relevance to them — because I believe this to be our group's key area of expertise.

  1. Random orthogonal projections were implemented in ortho_group as of scipy v0.18, as per this pull request; this discussion — to me at least — seems to highlight issues that arise when submitting a pull request to an involved and well-maintained operation.

  2. Functionality for Nearest Neighbour trees was expanded, as outlined by this blog article, in this pull request. Again, it seems to highlight the bare minimum of what must go into contributing tools to a major OSS operation.

As a side note, I believe each pull request highlights aspects of these algorithms that are not yet implemented, meaning they are something we could expand on! Even more exciting: there are significantly faster algorithms for sampling random matrices from O(n) than the (likely Householder-based) method used in the ortho_group code. This is also something we could improve!

asberk commented 7 years ago

I now realize this issue more rightly belongs in the 2017-Fall repo, if such a repo were to exist. I can move it when the iteration appears. 😄