Federated-EM with heterogeneity mitigation and variance reduction.
Aymeric DieuleveutGersende FortEric MoulinesGeneviève RobinPublished in: NeurIPS (2021)
Keyphrases
- variance reduction
- monte carlo
- gradient estimation
- sample size
- expectation maximization
- bias variance decomposition
- em algorithm
- random numbers
- importance sampling
- quasi monte carlo
- probabilistic model
- generative model
- maximum likelihood
- parameter estimation
- naive bayes classifier
- distributed systems
- confidence intervals
- learning algorithm
- supervised learning
- k means
- objective function
- training data