Learning No-Regret Sparse Generalized Linear Models with Varying Observation(s)

Authors: Diyang Li, Charles Ling, zhiqiang xu, Huan Xiong, Bin Gu

ICLR 2024 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental Encouraging results are exhibited on real-world benchmarks.
Researcher Affiliation Academia Diyang Li1, Charles X. Ling2, Zhiqiang Xu3, Huan Xiong3 & Bin Gu3, 1Cornell University 2Western University 3Mohamed bin Zayed University of Artificial Intelligence
Pseudocode Yes Algorithm 1 SAGO Algorithm
Open Source Code Yes To ensure the replicability, Python codes corresponding to the pivotal components of the proposed algorithms are incorporated within the supplementary materials.
Open Datasets Yes Dataset We employ real-world datasets from Open ML (Vanschoren et al., 2014) and UCI repository (Asuncion & Newman, 2007) for our simulations.
Dataset Splits Yes We randomly partition the datasets into training, validation, and testing sets, with 70%, 15%, and 15% of the total samples, respectively.
Hardware Specification Yes All experiments presented in this study were conducted on a workstation running the Ubuntu 18.04 operating system, equipped with Intel 2.30GHz CPU 200 and 400.0GB of RAM.
Software Dependencies No The paper mentions 'Python 3.7' as the implementation language and libraries like 'Num Py and Sci Py', 'Scikit-learn', and 'Hyperopt' without specifying version numbers for these libraries.
Experiment Setup Yes The parameterizers are set to µ (t) =4t2, ν (s) = 1 s2, respectively. ... The convergence tolerance ε for batch training is 1e-7 and the tolerance ϵ for hyperparameter in the outer-level problem is 1e-4.