Join GitHub today
GitHub is home to over 40 million developers working together to host and review code, manage projects, and build software together.Sign up
Lookahead Optimizer: k steps forward, 1 step back #138
This optimiser seems related (at least in spirit) to the Late Acceptance Hill-Climbing (LAHC) optimiser, which has been around since about 2008 in one form or another. A recent-ish journal article on LAHC is here: https://doi.org/10.1016/j.ejor.2016.07.012
Curiously, the arxiv pre-print for the Lookahead optimiser doesn't seem to mention LAHC at all. Maybe this is a case of convergent evolution.
rcurtin left a comment
The paper idea seems interesting and I enjoyed reading it, but I agree with @conradsnicta that it's related to things that are already out there. But I think that's the authors' problem and not so much ours. :)
The implementation looks correct to me---it's quite simple and easy to read both the paper and the code---but I do have a few comments about documentation and how we handle the inner optimizer type. Let me know what you think.