Robust Bayesian Regression via Hard Thresholding
Authors: Zheyi Fan, Zhaohui Li, Qingpei Hu
NeurIPS 2022 | Conference PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | Extensive experiments show that, under different dataset attacks, our algorithms achieve state-of-the-art results compared with other benchmark algorithms, demonstrating the robustness of the proposed approach. |
| Researcher Affiliation | Academia | 1Academy of Mathematics and Systems Science, Chinese Academy of Sciences, China 2School of Mathematical Sciences, University of Chinese Academy of Sciences, China 3H. Milton Stewart School of Industrial and Systems Engineering, Georgia Institute of Technology, USA. |
| Pseudocode | Yes | Algorithm 1 TRIP: hard Thresholding approach to Robust regression with s Imple Prior |
| Open Source Code | No | Did you include the code, data, and instructions needed to reproduce the main experimental results (either in the supplemental material or as a URL)? [No] The experiments are based on simulation and it is easy to reproduce under given parameters. |
| Open Datasets | No | In our experiments, the data generation can be divided into two steps. First, we generate the basic model. The true coefficient w is chosen to be a random unit norm vector. The covariant xi are iid in N(0, Id). The data are generated by yi = x T i w + ϵi, where ϵi are iid in N(0, σ2). We set σ = 1 in the experiments. |
| Dataset Splits | No | The paper describes data generation and corruption but does not specify training, validation, or test splits for the generated data. |
| Hardware Specification | No | Did you include the total amount of compute and the type of resources used (e.g., type of GPUs, internal cluster, or cloud provider)? [No] The overall calculation scale is small. |
| Software Dependencies | No | The paper does not provide specific software dependencies or version numbers used for implementation. |
| Experiment Setup | Yes | In our experiments, the data generation can be divided into two steps. First, we generate the basic model. The true coefficient w is chosen to be a random unit norm vector. The covariant xi are iid in N(0, Id). The data are generated by yi = x T i w + ϵi, where ϵi are iid in N(0, σ2). We set σ = 1 in the experiments. ... The prior coefficient w0 is generated by w + νu, where u is a random unit norm vector and ν is a non-negative number (ν is set to 0.5 unless otherwise stated). Σ0 takes the form s I, where s takes a different value for each method. All parameters are fixed in each experiment. ... δ is set to 0.1n for n = 1000, p = 200, and to 0.2n for n = 2000, p = 100. ... The prior distribution pr(r) is set to the Gamma distribution unless otherwise stated. |