Quite a nice neural network package that I'm starting to investigate: FANN.
Looks like it's been carefully optimized, has been around for a while, and is pretty easy to use. I've cored out the hand-rolled networks I was using before and am starting to look at performance benefits of using a proper optimized network package.
It also opens the door to more complex training algos than the standard backpropogation that I've been using so far, since those come in the FANN box. Probably most useful in the supervised learning version of training.
Looks like it's been carefully optimized, has been around for a while, and is pretty easy to use. I've cored out the hand-rolled networks I was using before and am starting to look at performance benefits of using a proper optimized network package.
It also opens the door to more complex training algos than the standard backpropogation that I've been using so far, since those come in the FANN box. Probably most useful in the supervised learning version of training.
No comments:
Post a Comment