Safe screening rules for L0-regression from Perspective Relaxations
Authors: Alper Atamturk, Andres Gomez
ICML 2020 | Conference PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | Numerical experiments on real and synthetic data indicate that a significant number of the variables can be removed quickly, hence reducing the computational burden for optimization substantially. Therefore, the proposed fast and effective screening rules extend the scope of algorithms for ℓ0-regression to larger data sets. |
| Researcher Affiliation | Academia | 1IEOR, University of California, Berkeley, USA 94720 2ISE, University of Southern California, Los Angeles, USA 90089. |
| Pseudocode | No | No structured pseudocode or algorithm blocks were found. |
| Open Source Code | No | The paper references code by Bertsimas et al. (2019) at https://github.com/jeanpauphilet/SubsetSelectionCIO.jl in a footnote, but does not state that the authors of this paper are releasing their own implementation code for the described methodology. |
| Open Datasets | Yes | The Diabetes data set is first used by Efron et al. (2004), whereas the other data sets are obtained from the UCI Machine Learning Repository (Dua & Graff, 2017). |
| Dataset Splits | No | The paper describes using synthetic and real datasets for computational experiments but does not explicitly provide details on how the data was split into training, validation, and test sets with specific percentages or counts for reproducibility. |
| Hardware Specification | Yes | All experiments are performed on a laptop with eight Intel(R) Core(TM) i7-8550 CPUs and 16GB RAM. |
| Software Dependencies | Yes | In our computations we use CPLEX 12.8 mixed-integer optimizer. |
| Experiment Setup | Yes | In our experiments, we let n = 1,000, m = 500, k {10, 30, 50}, γ = 2iγ0 with i { 1, 0, 2, 4} and γ0 = n mk maxi ai 2 2 (where ai denotes the i-th row of A), ρ {0.2, 0.5, 0.7}, and SNR {0.05, 1.00, 6.00}. The parameters m, γ, ρ and SNR coincide with the values used in Bertsimas et al. (2019). Our instances are smaller with n = 1, 000 and k {10, 30, 50} as we use a general purpose mixed-integer solver rather than a tailored solution method for (MIPC) as in Bertsimas et al. (2019). Several other papers in the literature generate data similarly. Finally, we set the time limit to ten minutes. |