Understanding Approximate Fisher Information for Fast Convergence of Natural Gradient Descent in Wide Neural Networks
–Neural Information Processing Systems
The fast convergence holds in layer-wise approximations; for instance, in block diagonal approximation where each block corresponds to a layer as well as in block tri-diagonal and K-FAC approximations.
Neural Information Processing Systems
Oct-3-2025, 08:08:43 GMT
- Country:
- Asia > Japan
- Honshū > Kantō > Tokyo Metropolis Prefecture > Tokyo (0.04)
- North America > Canada (0.04)
- Asia > Japan
- Genre:
- Research Report > New Finding (0.68)
- Technology: