Stochastic Online Linear Regression: the Forward Algorithm to Replace Ridge

Authors: Reda Ouhamma, Odalric-Ambrym Maillard, Vianney Perchet

NeurIPS 2021 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental Last, we provide numerical experiments to illustrate our results and endorse our intuitions.
Researcher Affiliation Collaboration Reda Ouhamma Univ. Lille, CNRS, Inria, Centrale Lille, UMR 9189 CRISt AL, F-59000 reda.ouhamma@univ-lille.fr Odalric. Maillard Univ. Lille, CNRS, Inria, Centrale Lille, UMR 9189 CRISt AL, F-59000 Vianney. Perchet Criteo, ENSAE, ENS PARIS-SACLAY
Pseudocode Yes Algorithm 1: Online ridge regression Algorithm 2: The forward algorithm Algorithm 3: OFULf algorithm
Open Source Code Yes Did you include the code, data, and instructions needed to reproduce the main experimental results (either in the supplemental material or as a URL)? [Yes] See Appendix G
Open Datasets No The paper uses internally generated synthetic data for experiments and does not provide access information for a publicly available or open dataset.
Dataset Splits No The paper describes online simulation experiments with generated data, and thus does not provide traditional train/validation/test dataset splits. It does not mention any explicit validation split.
Hardware Specification No The paper mentions that the type of resources used can be found in Appendix G, but no specific hardware details (e.g., GPU/CPU models, memory) are provided in the main text of the paper.
Software Dependencies No The paper does not provide specific ancillary software details with version numbers (e.g., library or solver names with versions) in the main text.
Experiment Setup Yes In Figures 2a and 2b we observe the effect of regularization on the performance of ridge and forward regressions in a 5-dimensional regression setting, we vary λ {1/T, 1/ log(T), 1, 10}, sample a zero mean Gaussian noise with σ = 0.1 and draw features uniformly from the unit ball. [...] We consider a 100-dimensional linear bandit with 10 arms, the parameter vector is drawn from the unit ball, actions are such that xt 200. Noise ϵt L= N(0, 10 1), λ = 10 5, δ = 10 3.