Zero-Regret Performative Prediction Under Inequality Constraints
Authors: Wenjing YAN, Xuanyu Cao
NeurIPS 2023 | Conference PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | Finally, we validate the effectiveness of our algorithm and theoretical results through numerical simulations. ... We conduct experiments on two examples: multi-task linear regression and multi-asset portfolio. The numerical results validate the effectiveness of our algorithm and theoretical analysis. Fig. 1 and Fig. 2 show the numerical results of the multi-task linear regression and the multi-asset portfolio, respectively. |
| Researcher Affiliation | Academia | Wenjing Yan Xuanyu Cao Department of Electronic and Computer Engineering The Hong Kong University of Science and Technology wj.yan@connect.ust.hk, eexcao@ust.hk |
| Pseudocode | Yes | Algorithm 1 Adaptive Primal-Dual Algorithm |
| Open Source Code | No | The paper does not provide any statement or link indicating that the source code for the described methodology is open or publicly available. |
| Open Datasets | No | The paper describes numerical experiments on 'multi-task linear regression' and 'multi-asset portfolio' problems. It mentions 'simulation details are provided in F of the supplementary file' but does not provide concrete access information (link, DOI, formal citation) to any specific publicly available dataset used for these simulations. |
| Dataset Splits | No | The paper conducts numerical simulations but does not specify exact training, validation, or test dataset splits (e.g., percentages or sample counts) for its experiments. It refers to 'simulation details' being in a supplementary file, but these details are not provided in the main paper. |
| Hardware Specification | No | The paper describes numerical experiments but does not provide any specific details about the hardware (e.g., CPU, GPU models, memory) used to run these experiments. |
| Software Dependencies | No | The paper mentions 'CVX: Matlab software for disciplined convex programming, version 2.1' in its references, which is a tool that could be used. However, it does not explicitly state that this software (or any other, with specific version numbers) was a dependency for their own experimental implementation. |
| Experiment Setup | Yes | We compare the proposed adaptive primal-dual algorithm (abbreviated as APDA) with two approaches. The first approach is PD-PS, which stands for the primal-dual (PD) algorithm used to find the performative stable (PS) points. The algorithm PD-PS is similar to APDA, but it uses only the first term in Eq. (8) as the approximate gradient. The second approach is baseline, which runs the same procedures as APDA with perfect knowledge of A, i.e., the performative effect is known. ... In both figures, we consider two settings for the sensitivity parameter of D(θ), namely ε = 1 and ε = 10. ... for the setting of T = 106. ... Set η = 1 T and ζt = 2 κ1(t 1)+2κ3 , t [T]. |