Additive Tree-Structured Conditional Parameter Spaces in Bayesian Optimization: A Novel Covariance Function and a Fast Implementation
This work addresses the challenge of sample-efficient optimization in conditional parameter spaces, which is incremental but improves flexibility and performance for applications like neural network compression and architecture search.
The paper tackles the problem of optimizing black-box functions with conditional parameter spaces by proposing an additive tree-structured covariance function and a parallel algorithm, resulting in significantly outperforming state-of-the-art methods like SMAC, TPE, and Jenatton et al. (2017) in benchmarks and neural network tasks.
Bayesian optimization (BO) is a sample-efficient global optimization algorithm for black-box functions which are expensive to evaluate. Existing literature on model based optimization in conditional parameter spaces are usually built on trees. In this work, we generalize the additive assumption to tree-structured functions and propose an additive tree-structured covariance function, showing improved sample-efficiency, wider applicability and greater flexibility. Furthermore, by incorporating the structure information of parameter spaces and the additive assumption in the BO loop, we develop a parallel algorithm to optimize the acquisition function and this optimization can be performed in a low dimensional space. We demonstrate our method on an optimization benchmark function, on a neural network compression problem, on pruning pre-trained VGG16 and ResNet50 models as well as on searching activation functions of ResNet20. Experimental results show our approach significantly outperforms the current state of the art for conditional parameter optimization including SMAC, TPE and Jenatton et al. (2017).