Speed it up. Possibly by finishing the multiprocessing version. Also maybe don't do this because once I've jammed a neural net in somewhere it could get awkward? Currently this is unclear.
Regardless the current version is rather slow, which makes it difficult to tell if it is even working so maybe getting the parallel one going for a baseline isn't a bad idea.
Speed it up. Possibly by finishing the multiprocessing version. Also maybe don't do this because once I've jammed a neural net in somewhere it could get awkward? Currently this is unclear.
Regardless the current version is rather slow, which makes it difficult to tell if it is even working so maybe getting the parallel one going for a baseline isn't a bad idea.