Robust Bayesian Satisficing
Authors: Artun Saday, Y. Cahit Yıldırım, Cem Tekin
NeurIPS 2023 | Conference PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | 5 Experiments We evaluate the proposed algorithm on one synthetic and one real-world environment. We compare the lenient regret and robust satisficing regret of Ro BOS with the following benchmark algorithms. ... Results for the robust satisficing and lenient regrets are given in Figure 2. ... As seen in Figures 4a and 4b, when the amount of distribution shift at each round is known exactly by DRBO, it can perform better than Ro BOS. However, when the distributional shift is either underestimated or overestimated, Ro BOS achieves better results. |
| Researcher Affiliation | Academia | Artun Saday Bilkent University artun.saday@bilkent.edu.tr, Ya sar Cahit Yıldırım Bilkent University cahit.yildirim@bilkent.edu.tr, Cem Tekin Bilkent University cemtekin@ee.bilkent.edu.tr |
| Pseudocode | Yes | Algorithm 1: Ro BOS |
| Open Source Code | Yes | Our implementation is available at http://github.com/Bilkent-CYBORG/Ro BOS. |
| Open Datasets | Yes | We use the open-source implementation of the U.S. FDA-approved University of Virginia (UVA)/PADOVA T1DM simulator [37, 5]. |
| Dataset Splits | No | The paper describes generating data through a simulator and specific distributions for parameters (e.g., 'ζt U(20, 80) and N is the random term setting the distributional shift. We ran our experiments with N U( 6, 6) and with Nt U( 6/ log(t+2), 6/ log(t+2))'), but it does not specify explicit training, validation, or test dataset splits in terms of percentages or sample counts. |
| Hardware Specification | No | The paper does not provide specific details about the hardware used for running experiments, such as CPU or GPU models, or cloud computing specifications. |
| Software Dependencies | No | The paper mentions various algorithms and models like GP-UCB, DRBO, Gaussian processes, and specific kernels, and refers to a simulator, but it does not specify version numbers for any of the software dependencies. |
| Experiment Setup | Yes | The paper provides specific experimental setup details such as 'The threshold τ is set to 60% of the maximum value of the objective function, i.e., τ = 0.6 max(x,c) X C f(x, c) 0.65Zt.', 'We use an RBF kernel with Automatic Relevance Determination (ARD) and lengthscales 0.2 and 5 respectively for dimensions X and C.', 'Simulations are run with observation noise ηt N(0, 0.022).', 'target blood glucose level as K = 112.5 mg/dl and define the pseudo-reward function as r(t) = |o(t) K|', 'setting the threshold τ = 10.', 'environment picks the true and reference distributions as Pt N(ζt, 2.25) and P t N(ζt + N, 9) where ζt U(20, 80) and N is the random term setting the distributional shift.', 'We ran our experiments with N U( 6, 6) and with Nt U( 6/ log(t+2), 6/ log(t+2)).', 'Simulations are run with observation noise ηt N(0, 1).', and 'For the GP surrogate, we used a Matérn-ν kernel with length-scale parameter 10.' |