Gradient Routing: Masking Gradients to Localize Computation in Neural Networks

Open in new window