Projection-free Distributed Online Convex Optimization with $O(\sqrtT)$ Communication Complexity

Authors: Yuanyu Wan, Wei-Wei Tu, Lijun Zhang

ICML 2020 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental In this section, we perform simulation experiments to verify the performance of our proposed algorithms. 5.1. Experimental Settings 5.2. Experimental Results
Researcher Affiliation Collaboration 1National Key Laboratory for Novel Software Technology, Nanjing University, Nanjing, China 24Paradigm Inc., Beijing, China.
Pseudocode Yes Algorithm 1 CGSC; Algorithm 2 D-BOCG; Algorithm 3 D-BBCG
Open Source Code No The paper does not provide concrete access to source code for the methodology described.
Open Datasets Yes As in Zhang et al. (2017), we also use the aloi dataset from the LIBSVM repository (Chang & Lin, 2011), the details of which are summarized in Table 1.
Dataset Splits No The paper does not provide specific dataset split information (e.g., percentages, sample counts, or citations to predefined splits) for training, validation, or testing.
Hardware Specification No No specific hardware details (e.g., GPU/CPU models, memory amounts) used for running the experiments are provided.
Software Dependencies No The paper does not provide specific ancillary software details with version numbers.
Experiment Setup Yes According to Zhang et al. (2017), we set the bound of trace norm as τ = 50, and set st = 1/ t and η = c T 3/4 for D-OCG by tunning the constant c. For our D-BOCG, we set K = T , ϵ = 1e 5, L = 20 and η = c T 3/4 by tunning the constant c. For both D-BOCG and D-OCG, the constant c is selected from [0.01, , 1e5]. ... For D-BBCG, we set K = T , ϵ = 1e 5, L = 20, δ = 0.1 and η = c T 3/4 by selecting the constant c from [0.01, , 1e5].