#differential geometry

Optimal Pruning for Neural Architectures Using FisherNEW⭐9

A Fisher information geodesic distance framework creates optimal neural network pruning methods that outperform magnitude and local Fisher pruning on MNIST and ...