AdaGrad stepsizes: Sharp convergence over nonconvex landscapes, from any initialization.
Rachel WardXiaoxia WuLéon BottouPublished in: CoRR (2018)
Keyphrases
- initial conditions
- initial guess
- globally convergent
- approximate dynamic programming
- objective function
- stationary points
- convergence rate
- global optimization
- high quality
- k means
- convex optimization
- machine learning
- convergence speed
- iterative algorithms
- optimization problems
- denoising
- dynamic programming
- evolutionary algorithm
- faster convergence
- search algorithm
- information systems
- genetic algorithm
- database