Local policy search with Bayesian optimization

Authors: Sarah Müller, Alexander von Rohr, Sebastian Trimpe

NeurIPS 2021 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental The comparison reveals improved sample complexity and reduced variance in extensive empirical evaluations on synthetic objectives. Further, we highlight the benefits of active sampling on popular RL benchmarks.
Researcher Affiliation Collaboration 1Max Planck Institute for Intelligent Systems, Stuttgart, Germany 2Institute for Data Science in Mechanical Engineering, RWTH Aachen University, Germany 3IAV Gmb H, Gifhorn, Germany 4 Institute for Ophthalmic Research, University of Tübingen, Tübingen, Germany
Pseudocode Yes Algorithm 1 GIBO
Open Source Code Yes All data and source code necessary to reproduce the results are published at https://github.com/sarmueller/gibo.
Open Datasets Yes Lastly, we evaluate the performance of GIBO on classic control and Mu Jo Co tasks included in the Open AI Gym [35, 36].
Dataset Splits No The paper describes using synthetic functions for 'within-model comparison' and RL environments (Gym and MuJoCo) for evaluation, showing 'training curves'. However, it does not provide explicit numerical train/validation/test splits (e.g., percentages or sample counts) or specify how data was partitioned for these purposes beyond general evaluation settings.
Hardware Specification No The paper does not specify any particular hardware components such as GPU or CPU models used for running the experiments.
Software Dependencies No The paper mentions software like 'Bo Torch' and 'Gpytorch' but does not specify their version numbers or any other software dependencies with version details.
Experiment Setup Yes All algorithms were started in the middle of the domain x0 = [0.5]d and had a limited budget of 300 noised function evaluations. The noise was Gaussian distributed with standard deviation σ = 0.1.