Home   Research Publications Members Related Software
IndexBrowse   BibliographiesMy selection
 Search: in   (word length ≥ 3)
      Login
Publication no #641   Download bibtex file Type :   Html | Bib | Both
Add to my selection
TrimTuner: Efficient Optimization of MachineLearning Jobs in the Cloud via Sub-Sampling

Pedro Mendes, Maria Casimiro, Paolo Romano and David Garlan.


In Proceedings of the 2020 Symposium on Modelling, Analysis, and Simulation of Computer and Telecommunication Systems (MASCOTS 2020), 2020.

Online links: PDF

Abstract
This work introduces TrimTuner, the first system for optimizing machine learning jobs in the cloud to exploit sub-sampling techniques to reduce the cost of the optimization process, while keeping into account user-specified constraints. TrimTuner jointly optimizes the cloud and application-specific parameters and, unlike state of the art works for cloud optimization, eschews the need to train the model with the full training set every time a new configuration is sampled. Indeed, by leveraging sub-sampling techniques and data-sets that are up to 60x smaller than the original one, we show that TrimTuner can reduce the cost of the optimization process by up to 50x. Further, TrimTuner speeds-up the recommendation process by 65×with respect to state of the art techniques for hyper-parameter optimization that use sub-sampling techniques. The reasons for this improvement are twofold: i) a novel domain specific heuristic that reduces the number of configurations for which the acquisition function has to be evaluated; ii) the adoption of an ensemble of decision trees that enables boosting the speed of the recommendation process by one additional order of magnitude.

Keywords: Machine Learning.  
@InProceedings{2020:MASCOTS,
      AUTHOR = {Mendes, Pedro and Casimiro, Maria and Romano, Paolo and Garlan, David},
      TITLE = {TrimTuner: Efficient Optimization of MachineLearning Jobs in the Cloud via Sub-Sampling},
      YEAR = {2020},
      BOOKTITLE = {Proceedings of the 2020 Symposium on Modelling, Analysis, and Simulation of Computer and Telecommunication Systems (MASCOTS 2020)},
      PDF = {http://acme.able.cs.cmu.edu/pubs/uploads/pdf/TrimTuner-MASCOTS-20.pdf},
      ABSTRACT = {This work introduces TrimTuner, the first system for optimizing machine learning jobs in the cloud to exploit sub-sampling techniques to reduce the cost of the optimization process, while keeping into account user-specified constraints. TrimTuner jointly optimizes the cloud and application-specific parameters and, unlike state of the art works for cloud optimization, eschews the need to train the model with the full training set every time a new configuration is sampled. Indeed, by leveraging sub-sampling techniques and data-sets that are up to 60x smaller than the original one, we show that TrimTuner can reduce the cost of the optimization process by up to 50x. Further, TrimTuner speeds-up the recommendation process by 65×with respect to state of the art techniques for hyper-parameter optimization that use sub-sampling techniques. The reasons for this improvement are twofold: i) a novel domain specific heuristic that reduces the number of configurations for which the acquisition function has to be evaluated; ii) the adoption of an ensemble of decision trees that enables boosting the speed of the recommendation process by one additional order of magnitude.},
      KEYWORDS = {Machine Learning}
}
    Created: 2021-02-12 18:01:06     Modified: 2021-08-02 15:47:12
Feedback: ABLE Webmaster
Last modified: Sat October 12 2019 16:15:32
        BibAdmin