Blind Descent: A Prequel to Gradient Descent

Akshat Gupta, Prasad N R

We describe an alternative learning method for neural networks, which we call Blind Descent. By design, Blind Descent does not face problems like exploding or vanishing gradients.…