Hyperparameter Tuning in Bandit-Based Adaptive Operator Selection
We are using bandit-based adaptive operator selection while autotuning parallel computer programs. The autotuning, which uses evolutionary algorithm-based stochastic sampling, takes place over an extended duration and occurs in situ as programs execute. The environment or context during tuning is either largely static in one scenario or dynamic in another. We rely upon adaptive operator selection to dynamically generate worthy test configurations of the program. In this paper, we study how the choice of hyperparameters, which control the trade-off between exploration and exploitation, affects the effectiveness of adaptive operator selection which in turn affects the performance of the autotuner. We show that while the optimal assignment of hyperparameters varies greatly between different benchmarks, there exists a single assignment, for a context, of hyperparameters that performs well regardless of the program being tuned.
KeywordsScore Function Mutation Operator Image Compression Convergence Time Dynamic Context
Unable to display preview. Download preview PDF.
- 1.Ansel, J., Chan, C., Wong, Y.L., Olszewski, M., Zhao, Q., Edelman, A., Amarasinghe, S.: Petabricks: A language and compiler for algorithmic choice. In: ACM SIGPLAN Conference on Programming Language Design and Implementation, Dublin, Ireland (June 2009)Google Scholar
- 2.Ansel, J., Wong, Y.L., Chan, C., Olszewski, M., Edelman, A., Amarasinghe, S.: Language and compiler support for auto-tuning variable-accuracy algorithms. In: International Symposium on Code Generation and Optimization, Chamonix, France (April 2011)Google Scholar
- 5.Fialho, Á.: Adaptive Operator Selection for Optimization. PhD thesis, Université Paris-Sud XI, Orsay, France (December 2010)Google Scholar
- 6.Fialho, Á., Da Costa, L., Schoenauer, M., Sebag, M.: Analyzing bandit-based adaptive operator selection mechanisms. Annals of Mathematics and Artificial Intelligence – Special Issue on Learning and Intelligent Optimization (2010)Google Scholar
- 8.Maturana, J., Fialho, Á., Saubion, F., Schoenauer, M., Sebag, M.: Extreme com-pass and dynamic multi-armed bandits for adaptive operator selection. In: CEC 2009: Proc. IEEE International Conference on Evolutionary Computation, pp. 365–372. IEEE Press (May 2009)Google Scholar
- 9.Pacula, M.: Evolutionary algorithms for compiler-enabled program autotuning. Master’s thesis, Massachusetts Institute of Technology, Cambridge, MA (2011)Google Scholar