Gradient Descent on Neural Networks Typically Occurs at the Edge of Stability.
Jeremy M. CohenSimran KaurYuanzhi LiJ. Zico KolterAmeet TalwalkarPublished in: CoRR (2021)
Keyphrases
- neural network
- cost function
- learning rules
- loss function
- pattern recognition
- genetic algorithm
- neural nets
- objective function
- fuzzy logic
- edge detection
- stability analysis
- fuzzy neural network
- lyapunov function
- machine learning
- weighted graph
- edge detector
- data sets
- pairwise
- neural network model
- self organizing maps
- recurrent neural networks
- fault diagnosis
- multi layer
- back propagation
- learning algorithm
- error function
- image segmentation