Movement Penalized Bayesian Optimization with Application to Wind Energy Systems
Authors: Shyam Sundhar Ramesh, Pier Giuseppe Sessa, Andreas Krause, Ilija Bogunovic
NeurIPS 2022 | Conference PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | We further demonstrate our approach on the important real-world application of altitude optimization for Airborne Wind Energy Systems. In the presence of substantial movement costs, our algorithm consistently outperforms standard CBO algorithms. and This section provides numerical results on synthetic and real-world data. |
| Researcher Affiliation | Academia | Shyam Sundhar Ramesh ETH Zurich, Pier Giuseppe Sessa ETH Zurich, Andreas Krause ETH Zurich, Ilija Bogunovic University College London |
| Pseudocode | Yes | Algorithm 1 GP-MD |
| Open Source Code | Yes | Added in supplementary material. |
| Open Datasets | Yes | In this section, we use a dataset from [9] which contains wind-speed information over various locations in Europe for a period ranging from 2011 to 2017, and also includes measurements at different altitudes per location. and Philip Bechtle, Mark Schelbergen, Roland Schmehl, Udo Zillmann, and Simon Watson. Air-borne wind energy resource analysis. Renewable energy, 2019. |
| Dataset Splits | No | The paper describes sampling strategies and iterations for experiments, but does not explicitly provide training/test/validation dataset splits (e.g., percentages or sample counts) needed to reproduce data partitioning. |
| Hardware Specification | No | Internal Cluster was used - this is too vague and lacks specific hardware details such as GPU/CPU models or memory. |
| Software Dependencies | No | The paper does not explicitly provide specific software names with version numbers for reproducibility. |
| Experiment Setup | Yes | We use the same constant value β = 2.0 for the exploration parameter in both GP-MD and CGP-LCB (since the theoretical worst-case bounds are found to be overly pessimistic and can impede performance [50]). and To learn Vw(x, t), we normalize the inputs, and fit a GP with RBF kernel (lengthscale=3.67, outputscale=6.85 and noise parameter=2.73). |