quinngroup / fergus-ssl

Python implementation of the Rob Fergus semi-supervised learning algorithm.

Geek Repo:Geek Repo

Github PK Tool:Github PK Tool

Performance testing of Fergus SSL

magsol opened this issue · comments

Need a performance benchpark for the current algorithmic framework. Use sklearn.datasets to conduct semi-supervised learning testing, and time the results for varying dataset sizes as well as accuracy.

We still need timings for the dataset.

  • How long does it take to run?
  • How does changing the size of the dataset change the runtime?
  • Where are the bottlenecks?
  • How does varying gamma affect the clustering quality?
  • How does adding noise to the half-moons affect the clustering quality?