After a little busy while, I have graduated and entered the summer vacation, which means time for serious GSoC work.
So we had a little conference in order to discuss what will be done and when. We gathered quite a few code snippets since the official start of the project, but it’s now time to talk about integration and pull requests.
Here is the plan:
First pull request due: June 15
This will be the use case I blogged about before. Specifically, we want to learn a dictionary of sparse atoms, but representations of the data will be dense.
First pull request due: June 25
This is the transpose of the SparsePCA problem. We are learning the optimal, dense dictionary for sparse representations of the data.
First pull request due: June 30
This method builds the dictionary out of cluster centers found by K-means.
First pull request due: July 10
This will involve the online learning tricks suggested in Julien Mairal’s work and will allow for faster computations of both sparse PCA and sparse coding. In the case of sparse coding, it will make use of the scikits.learn API for online learning.
While I will try to keep the deadlines for the initial pull requests as strictly as I can, we did not establish deadlines for merging, since this will depend on more factors. As long as the pull requests are up, the code review system will push it forward towards the merge. The focus is on teamwork and on feedback cycles as short as possible, as opposed to falling into the trap of delaying work until the night before the deadline.