ReinBo: Machine Learning Pipeline Conditional Hierarchy Search and Configuration with Bayesian Optimization Embedded Reinforcement Learning
ReinBo: Machine Learning Pipeline Conditional Hierarchy Search and Configuration with Bayesian Optimization Embedded Reinforcement Learning
复制标题
DOI：
10.1007/978-3-030-43823-4_7
复制
发表时间：
2019-09
期刊：
影响因子：
--
通讯作者：
Xudong Sun;Jiali Lin;B. Bischl
中科院分区：
文献类型：
--
作者：
Xudong Sun;Jiali Lin;B. Bischl
机器学习管道可能包括几个阶段的操作，如数据预处理、特征工程和机器学习模型训练。每个操作都有一组超参数，当未选择该操作时，这些参数可能与管线无关。这产生了分层条件超参数空间。为了优化这种混合的连续和离散条件分层超参数空间，我们提出了一种结合强化学习和贝叶斯优化能力的高效流水线搜索和配置算法。实验结果表明，与自动滑行、TPOT、Tree Parzen窗口和随机搜索等方法相比，本文提出的方法具有更好的性能。
Machine learning pipeline potentially consists of several stages of operations like data preprocessing, feature engineering and machine learning model training. Each operation has a set of hyper-parameters, which can become irrelevant for the pipeline when the operation is not selected. This gives rise to a hierarchical conditional hyper-parameter space. To optimize this mixed continuous and discrete conditional hierarchical hyper-parameter space, we propose an efficient pipeline search and configuration algorithm which combines the power of Reinforcement Learning and Bayesian Optimization. Empirical results show that our method performs favorably compared to state of the art methods like Auto-sklearn, TPOT, Tree Parzen Window, and Random Search.