Double Variance Reduction: A Smoothing Trick for Composite Optimization Problems without First-Order Gradient.
Hao DiHaishan YeYueling ZhangXiangyu ChangGuang DaiIvor W. TsangPublished in: CoRR (2024)
Keyphrases
- variance reduction
- optimization problems
- gradient estimation
- policy gradient
- evolutionary algorithm
- sample size
- monte carlo
- higher order
- objective function
- bias variance decomposition
- importance sampling
- random numbers
- quasi monte carlo
- naive bayes classifier
- confidence intervals
- trade off
- lower bound
- computational complexity
- genetic algorithm
- machine learning