Redis backend for Brain?
christopherdebeer opened this issue · 5 comments
Is anyone interested-in or working-on a Redis backend for this project?
Is there a reason there isn't one already (something I may be overlooking), or any fundamental reasons this wouldn't work?
Yeah, there's a reason there isn't one. The network can't be trained incrementally, you can only call train() once with all the data, so there's no point in keeping the network state around in Redis.
You can however collect the data incrementally, but that could be done outside the library and fed into train() once you've collected enough. If you just want to store the fully trained network in Redis, you can call network.toJSON() and put it in yourself.
I wasn't aware that a network couldnt be trained more than once - damn :( my bad. thanks for the reply.
Sorry quick question, after I posted my previous comment, i tried it out (training an already trained network) and it works (in so far as it doesn't throw any errors) Is this essentially re-training the network from scratch using the new data, or is it just screwing the network up?
That's the same as retraining from scratch. Unfortunately, the best way to deal with this is to build up the data, and re-train periodically with everything you've collected so far.
Thanks :) clarification much appreciated.