Projection-free Distributed Online Learning in Networks
Authors: Wenpeng Zhang, Peilin Zhao, Wenwu Zhu, Steven C. H. Hoi, Tong Zhang
ICML 2017 | Conference PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | Experiments on two large-scale real-world datasets for a multiclass classification task confirm the computational benefit of the proposed algorithm and also verify the theoretical regret bound. |
| Researcher Affiliation | Collaboration | 1Department of Computer Science and Technology, Tsinghua University, Beijing, China 2Artificial Intelligence Department, Ant Financial Services Group, Hangzhou, China 3School of Information Systems, Singapore Management University, Singapore 4Tencent AI Lab, Shenzhen, China. |
| Pseudocode | Yes | Algorithm 3 Distributed Online Conditional Gradient (D-OCG) |
| Open Source Code | No | The paper does not provide any explicit statements about releasing source code or links to a code repository for the described methodology. |
| Open Datasets | Yes | We use two multiclass datasets selected from the LIBSVM2 repository with relatively large number of instances, which is summarized in Table 1. ... 2https://www.csie.ntu.edu.tw/~cjlin/libsvmtools/datasets/ |
| Dataset Splits | No | The paper describes an 'online learning setting' where data arrives sequentially over T rounds and doesn't specify traditional train/validation/test splits with percentages or counts for reproducibility. |
| Hardware Specification | No | The paper does not provide specific details regarding the hardware used for conducting the experiments. |
| Software Dependencies | No | The paper mentions parameter settings and methods but does not specify any software dependencies with version numbers (e.g., programming languages, libraries, frameworks). |
| Experiment Setup | Yes | Parameter Settings We set most of the parameters in these algorithms as what their corresponding theories suggest. For instance, the parameters σt,i in D-OCG are strictly set to be 1/t and the learning rates in D-OGD are set to be the typical decaying sequence 1/t. We use the method utilized in (Duchi et al., 2012) to generate the doubly stochastic matrices and fix the nuclear norm bound τ to 50 throughout. |