Maximum Likelihood-based Online Adaptation of Hyper-parameters in CMA-ES.
Ilya LoshchilovMarc SchoenauerMichèle SebagNikolaus HansenPublished in: CoRR (2014)
Keyphrases
- hyperparameters
- maximum likelihood
- model selection
- em algorithm
- maximum a posteriori
- cross validation
- gaussian process
- bayesian inference
- closed form
- bayesian framework
- support vector
- random sampling
- variational bayes
- noise level
- evolution strategy
- cma es
- prior information
- incremental learning
- sample size
- expectation maximization
- incomplete data
- parameter settings
- parameter space
- covariance matrix
- markov random field
- active learning
- machine learning
- probabilistic model
- evolutionary algorithm
- training data