Completed
Two-Layer Neural Network
Class Central Classrooms beta
YouTube videos curated by Class Central.
Classroom Contents
Beyond Lazy Training for Over-parameterized Tensor Decomposition
Automatically move to the next video in the Classroom when playback concludes
- 1 Intro
- 2 Tensor (CP) decomposition
- 3 Why naïve algorithm fails
- 4 Why gradient descent?
- 5 Two-Layer Neural Network
- 6 Form of the objective
- 7 Difficulties of analyzing gradient descent
- 8 Lazy training fails
- 9 O is a high order saddle point
- 10 Our (high level) algorithm
- 11 Proof ideas
- 12 Iterates remain close to correct subspace
- 13 Escaping local minima by random correlation
- 14 Amplify initial correlation by tensor power method
- 15 Conclusions and Open Problems
