Adaptive Algorithms for Online Convex Optimization with Long-term Constraints

Authors: Rodolphe Jenatton, Jim Huang, Cedric Archambeau

ICML 2016 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental We supplement the analysis with experiments validating the performance of our algorithm in practice.We ran two sets of experiments to assess the performance for our adaptive algorithms for OCO with long-term constraints and compare to the algorithms proposed by Mahdavi et al. (2012a).
Researcher Affiliation Industry Rodolphe Jenatton , Jim C. Huang , Cedric Archambeau {JENATTON,HUANGJIM,CEDRICA}@AMAZON.COM Amazon, Berlin, Germany, Seattle, USA
Pseudocode Yes Initialize x1 = 0 and λ1 = 0. For t {1, , T 1}: xt+1 = ΠB(xt ηt x Lt(xt, λt)), λt+1 = ΠR+(λt + µt λLt(xt, λt)),
Open Source Code No The paper does not provide any statement or link indicating that the source code for their methodology is publicly available.
Open Datasets Yes We solve the above problem using the datasets ijcnn1 and covtype, consisting respectively of 49, 990 and 581, 012 samples of dimension d = 22 and d = 54 each.2 www.csie.ntu.edu.tw/ cjlin/libsvmtools/datasets/binary.html.
Dataset Splits No The paper describes using datasets 'ijcnn1' and 'covtype' and discusses generating sequences, but it does not specify explicit train/validation/test splits, percentages, or sample counts for these datasets.
Hardware Specification No The paper does not provide any specific details about the hardware used to run the experiments, such as CPU/GPU models or memory specifications.
Software Dependencies No The paper mentions using 'CVXPY' and an implementation based on 'Defazio et al. (2014)' but does not provide specific version numbers for these or other software dependencies.
Experiment Setup Yes The parameter ρ is set to obtain approximately 30% of nonzero variables. They are computed over T = 1000 iterations with d = 64, and are averaged over 10 random sequences {Yt}T t=1.