Random Walk Initialization for Training Very Deep Feedforward Networks

Open in new window