Bayesian approach in neural networks for model pruning
Intro
Using deep learning in solving complex, real-world problems has become quite an engineering routine. But we should never forget about the probabilistic sense inside models and loss minimization. So here, we are going to recall that probabilistic framework and extend it to the Bayesian framework. Such switching will give us pleasant perks but it is not always for free. The main applicition presented in the blog will be neural networks pruning but other directions will be mentioned too.
So we will present 4 Bayesian techniques to envelope any task involving neural networks. If you get interested in the topic and want to use it on practice, you are welcome at our pytoch library.