Program Search for Machine Learning Pipelines Leveraging Symbolic Planning and Reinforcement Learning

ReinBo: Machine Learning Pipeline Conditional Hierarchy Search and Configuration with Bayesian Optimization Embedded Reinforcement Learning

“…Other Reinforcement Learning based methods In [32], the authors also combine pipeline search and hyper-parameter optimization in a reinforcement learning process based on the PEORL [33] framework, however, the hyperparameter is randomly sampled during the reinforcement learning process, an extra stage is needed to sweep the hyper-parameters using hyper-parameter optimization techniques, while in our work, hyper-parameter optimization is embedded in the reinforcement learning process. Alpha3M [14] combined MCTS and recurrent neural network in a self play [27] fashion, however, it seems that Alpha3M does not perform better than the state of art AutoML systems.…”

“…-To our best knowledge, we are the first to embed Bayesian Optimization (BO) into Reinforcement learning, specifically Q Learning [31] for collaborative joint search of pipelines and hyper-parameters, which is different from using BO for policy optimization [12], and also different from using BO for hyper-parameter fine tuning after an optimal pipeline is selected by a reinforcement learning based AutoML framework [32]. -We provide an open source light weight R language implementation reinbo 1 for the R Machine Learning community which could run efficiently on a personal computer, and takes much less resources compared to other Au-toML softwares.…”

Verifiably Safe Exploration for End-to-End Reinforcement Learning

“…While many methods extract a symbolic mapping for RL from visual data, e.g. (Lyu et al, 2019;Yang et al, 2018Yang et al, , 2019Lu et al, 2018;Garnelo et al, 2016;Li et al, 2018;Liang & Boularias, 2018;Goel et al, 2018), they all require that all of the reward-relevant features are explicitly represented in the symbolic space. As shown by the many successes of Deep RL, e.g.…”