On the Learnability of Deep Random Networks
This addresses the fundamental challenge of training deep networks for researchers in machine learning, though it appears incremental as it builds on existing theoretical and practical studies.
The paper tackles the problem of learnability in deep random networks, showing theoretically that learnability with sign activation drops exponentially with depth, and practically confirming this sharp drop with state-of-the-art training methods.
In this paper we study the learnability of deep random networks from both theoretical and practical points of view. On the theoretical front, we show that the learnability of random deep networks with sign activation drops exponentially with its depth. On the practical front, we find that the learnability drops sharply with depth even with the state-of-the-art training methods, suggesting that our stylized theoretical results are closer to reality.