4 ms·
I wander how many people still implement their own networks as opposed to use these prepared frameworks. Or do you guys stick to single framework or use some so
by zk00006 10y ago
I wander how many people still implement their own networks as opposed to use these prepared frameworks. Or do you guys stick to single framework or use some sort of mixture of tools?
- toxik 10y agoImplementing a basic neural network is a lot of work, to be honest. I tried in C, in fact even made it parallelizable. It was hell. Lots of hair-pulling. I thank the gods for Keras and keep my head down.
- Joof 10y agoUnsurprising. Normally, you at least have something like BLAS to do linear algebra work. It takes maybe 15 minutes in Python / numpy for something basic like a 2-layer net, but backpropegation is a little annoying to get right. Thus, tensorflow (or theano or caffe or whatever).
- Eridrus 10y agoOnly as a learning exercise, for anything you actually care about results for you want a GPU implementation.