Optimizing deep learning hyper-parameters through an evolutionary algorithm.
Steven R. YoungDerek C. RoseThomas P. KarnowskiSeung-Hwan LimRobert M. PattonPublished in: MLHPC@SC (2015)
Keyphrases
- deep learning
- hyperparameters
- evolutionary algorithm
- model selection
- cross validation
- closed form
- unsupervised learning
- bayesian framework
- bayesian inference
- support vector
- em algorithm
- prior information
- random sampling
- machine learning
- noise level
- sample size
- maximum likelihood
- incomplete data
- maximum a posteriori
- incremental learning
- fitness function
- genetic algorithm
- parameter settings
- weakly supervised
- expectation maximization
- parameter space
- missing values
- text classification
- probabilistic model
- feature selection
- computer vision
- mental models
- object recognition
- reinforcement learning
- object detection
- prior knowledge
- feature extraction
- image processing
- image reconstruction