1 paper · 1 filter
Yotam Alexander, Nimrod De La Vega, Noam Razin +1
The question of what makes a data distribution suitable for deep learning is a fundamental open problem. Focusing on locally connected neural networks (a prevalent family of archit…