A Universal Catalyst for First-Order Optimization

Authors: Hongzhou Lin, Julien Mairal, Zaid Harchaoui

NeurIPS 2015 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental We evaluate the Catalyst acceleration on three methods that have never been accelerated in the past: SAG [24], SAGA [6], and MISO-Prox. We focus on ℓ2-regularized logistic regression, where the regularization parameter µ yields a lower bound on the strong convexity parameter of the problem. We use three datasets used in [14], namely real-sim, rcv1, and ocr, which are relatively large, with up to n = 2 500 000 points for ocr and p = 47 152 variables for rcv1.
Researcher Affiliation Academia Hongzhou Lin1, Julien Mairal1 and Zaid Harchaoui1,2 1Inria 2NYU {hongzhou.lin,julien.mairal}@inria.fr zaid.harchaoui@nyu.edu
Pseudocode Yes Algorithm 1 Catalyst input initial estimate x0 Rp, parameters κ and α0, sequence (εk)k 0, optimization method M;
Open Source Code No The paper does not contain any statements about releasing open-source code for the methodology or provide links to a code repository.
Open Datasets Yes We use three datasets used in [14], namely real-sim, rcv1, and ocr, which are relatively large, with up to n = 2 500 000 points for ocr and p = 47 152 variables for rcv1.
Dataset Splits No The paper states it uses datasets 'real-sim, rcv1, and ocr' but does not specify any training, validation, or test splits (e.g., '80/10/10 split' or specific sample counts for each).
Hardware Specification No The paper does not provide specific details about the hardware used to run the experiments, such as exact GPU or CPU models, or memory specifications.
Software Dependencies No The paper mentions 'Python' in Appendix E regarding implementation details but does not provide specific version numbers for Python or any other key software libraries or solvers used in the experiments.
Experiment Setup Yes We compare MISO, SAG, and SAGA with their default parameters, which are recommended by their theoretical analysis (step-sizes 1/L for SAG and 1/3L for SAGA), and study several accelerated variants. The values of κ and ρ and the sequences (εk)k 0 are those suggested in the previous sections, with η=0.1 in (10). Other implementation details are presented in Appendix E.