Blog

Autograd, Fixed Points, and Tridiagonal Solvers

What happens if you parallelise backpropagation across layers?