Login / Signup
Structured Inverse-Free Natural Gradient: Memory-Efficient & Numerically-Stable KFAC for Large Neural Nets.
Wu Lin
Felix Dangel
Runa Eschenhagen
Kirill Neklyudov
Agustinus Kristiadi
Richard E. Turner
Alireza Makhzani
Published in:
CoRR (2023)
Keyphrases
</>
neural nets
memory efficient
numerically stable
natural gradient
feed forward
independent component analysis
blind source separation
back propagation
numerical stability
learning tasks
learning rate
polynomial equations
neural network
artificial neural networks
fixed point
activation function
pairwise