Join GitHub today
GitHub is home to over 50 million developers working together to host and review code, manage projects, and build software together.Sign up
GitHub is where the world builds software
Millions of developers and companies build, ship, and maintain their software on GitHub — the largest and most advanced development platform in the world.
[MRG+1] GaussianProcessRegressor: faster prediction of std #8591
What does this implement/fix? Explain your changes.
Predicting the std in GaussianProcessRegressor is very slow because of a particular np.einsum call. Simplifying the call gives the same result much faster.
Any other comments?
See this SO thread for a very similar situation solved the same way:
See this gist for benchmarking:
Even for medium size matrix, the new call is over 10x faster.