K-FAC
Description: Kronecker-factored approximate curvature is an optimization method for neural networks that approximates second-order information to compute parameter updates.
Primary topic: Stochastic Gradient Optimization Techniques (57%)
OpenAlex ID: keywords/k-fac
