Optimizing Conditional Value-At-Risk of Black-Box Functions
Authors: Quoc Phong Nguyen, Zhongxiang Dai, Bryan Kian Hsiang Low, Patrick Jaillet
NeurIPS 2021 | Conference PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | The performances of both CV-UCB and CV-TS are empirically evaluated in optimizing CVa R of synthetic benchmark functions and simulated real-world optimization problems. |
| Researcher Affiliation | Academia | Dept. of Computer Science, National University of Singapore, Republic of Singapore Dept. of Electrical Engineering and Computer Science, MIT, USA |
| Pseudocode | Yes | Algorithm 1 BO Algorithms for optimizing CVa R of a black-box function |
| Open Source Code | Yes | 2The code is available at https://github.com/qphong/BayesOpt-LV. |
| Open Datasets | Yes | We empirically evaluate the performance of both CV-UCB and CV-TS in optimizing CVa R of synthetic benchmark functions, an optimization problem of the residuary resistance per unit weight of displacement of a yacht, a portfolio optimization problem, and a simulated robot task (Sec. 5). In the experiment with the yacht hydrodynamics dataset, we would like to minimize the residuary resistance per unit weight of displacement of a yacht by searching for the optimal hull geometry coefficients of the yacht in the face of the uncertainty in the Froude number (the Froude number depends on the real-world environment and we assume that it can be simulated with computers during the optimization). The ground truth function is constructed using the yacht hydrodynamics data set [5]. |
| Dataset Splits | No | The paper describes experiments and mentions initial observations, but does not specify explicit training, validation, or test dataset splits (e.g., percentages or sample counts). The nature of Bayesian Optimization means data is gathered iteratively, which might not use fixed splits in the traditional sense, but the reproducibility requirements for splits are not met. |
| Hardware Specification | No | The paper does not provide specific details about the hardware (e.g., CPU, GPU models, memory) used for running the experiments. |
| Software Dependencies | No | The paper does not provide specific software dependencies with version numbers (e.g., Python, PyTorch, TensorFlow versions). |
| Experiment Setup | Yes | The risk level α is set to 0.1. Each experiment is repeated 10 times with different random seeds to account for the randomness in the observation noise, the set of initial observations, and the sampling from the GP posterior belief. |