Stochastic Sign Descent Methods: New Algorithms and Better Theory.
Mher SafaryanPeter RichtárikPublished in: ICML (2021)
Keyphrases
- significant improvement
- computational cost
- high computational complexity
- methods require
- methods outperform
- benchmark datasets
- heuristic methods
- algorithms require
- computationally expensive
- mathematical models
- exhaustive search
- optimization methods
- complexity analysis
- computer vision algorithms
- problems in computer vision
- computationally intensive
- synthetic and real data sets
- computational complexity
- search methods
- learning algorithm
- theoretical analysis
- optimization problems
- stochastic methods
- monte carlo methods
- methods can be applied
- theoretical models
- mathematical foundations
- experimental comparison
- preprocessing
- quality measures
- machine learning algorithms
- linear algebra
- theoretical framework
- monte carlo
- convergence rate
- learning models