[mlpack] Neural Evolution Algorithms - Week 9

[bang liu]

Dear all,

Last week, we continue working on finding out bugs in our code to try to
make the neat pass the Cart Pole balancing problem. We revised a few
important bugs (thanks very much to Marcus, gives me a lot help on
debugging.) We also tested with different parameter settings to check the
performance in Cart Pole balancing task, to try to reveal the underline
problems. But still the Cart Pole balancing test not success.

We will continue debugging to try to make the code pass the test in this
week. My current idea is incorporating other operations during the
Reproduce function in NEAT, by referring to the original NEAT
implementation code. Besides, more importantly, the key problem is still
figure our why the fitness on original Cart Pole balancing problem doesn't
decrease after some iterations. If this is not clear, maybe even more
operations won't help too much. Hope we could figure out why this week.

Best,
Bang
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mailman.cc.gatech.edu/pipermail/mlpack/attachments/20160718/caa58f6e/attachment-0002.html>