AI-powered
podcast player
Listen to all your favourite podcasts with AI-powered features
The Advantages of Quantization in Binary Neural Networks
In binary neural networks are this super interesting active area of research that's been around for quite some time. In many cases compress a model, say like 50%, almost no drop in performance. But there are really great intermediary states where you press it quite a bit and still maintain a lot of performance. This points to I think one of the things that makes this sort of new software 2.0 style computing paradigm so interesting is that your compute resources are very elastic.