Monte Carlo Tree Search based Variable Selection for High Dimensional Bayesian Optimization
About
Bayesian optimization (BO) is a class of popular methods for expensive black-box optimization, and has been widely applied to many scenarios. However, BO suffers from the curse of dimensionality, and scaling it to high-dimensional problems is still a challenge. In this paper, we propose a variable selection method MCTS-VS based on Monte Carlo tree search (MCTS), to iteratively select and optimize a subset of variables. That is, MCTS-VS constructs a low-dimensional subspace via MCTS and optimizes in the subspace with any BO algorithm. We give a theoretical analysis of the general variable selection method to reveal how it can work. Experiments on high-dimensional synthetic functions and real-world problems (i.e., NAS-bench problems and MuJoCo locomotion tasks) show that MCTS-VS equipped with a proper BO optimizer can achieve state-of-the-art performance.
Related benchmarks
| Task | Dataset | Result | Rank | |
|---|---|---|---|---|
| High-dimensional optimization | MSLR | Convergence Value-8.8755 | 21 | |
| High-dimensional optimization | Lasso-Hard | Convergence Value9.4854 | 20 | |
| High-dimensional optimization | LIMO | Convergence Value-5.3277 | 20 | |
| Function Optimization | Sphere D=1000 | Final Value36.5522 | 19 | |
| Function Optimization | Dixon D=1000 | Convergence Value1.27e+5 | 19 | |
| Function Optimization | Levy D=1000 | Convergence Value32.6922 | 19 | |
| Function Optimization | Rosenbrock D=1000 | Convergence Value1.29e+5 | 19 | |
| Function Optimization | Michalewicz D=1000 | Convergence Value-7.9805 | 19 | |
| Function Optimization | Griewank D=1000 | Convergence Value (Statistic)62.4722 | 19 | |
| High-dimensional optimization | Michalewicz D=10000 | Convergence Value-10.7175 | 13 |