Provable Methods for Training Neural Networks with Sparse Connectivity
This work addresses the challenge of sparse connectivity in neural networks for researchers and practitioners, offering a provable method that is incremental by building on existing linear network techniques.
The paper tackles the problem of training feedforward neural networks with sparse connectivity by providing novel guaranteed approaches, leveraging techniques from linear networks to learn non-linear networks through moment factorization, and shows that the output can serve as effective initializers for gradient descent.
We provide novel guaranteed approaches for training feedforward neural networks with sparse connectivity. We leverage on the techniques developed previously for learning linear networks and show that they can also be effectively adopted to learn non-linear networks. We operate on the moments involving label and the score function of the input, and show that their factorization provably yields the weight matrix of the first layer of a deep network under mild conditions. In practice, the output of our method can be employed as effective initializers for gradient descent.