RubberBand
Ujval Misra, Richard Liaw, Lisa Dunlap, Romil Bhardwaj, Kirthevasan Kandasamy, Joseph E. Gonzalez, Ion Stoica, Alexey Tumanov
Abstract
Hyperparameter tuning is essential to achieving state-of-theart accuracy in machine learning (ML), but requires substantial compute resources to perform. Existing systems primarily focus on eectively allocating resources for a hyperparameter tuning job under xed resource constraints. We show that the available parallelism in such jobs changes dynamically over the course of execution and, therefore, presents an opportunity to leverage the elasticity of the cloud.
Topics & Concepts
Leverage (statistics)Computer scienceHyperparameterCloud computingElasticity (physics)Distributed computingMulti-core processorMachine learningArtificial intelligenceParallel computingOperating systemComposite materialMaterials scienceParallel Computing and Optimization TechniquesMachine Learning and Data ClassificationCloud Computing and Resource Management