Federated Latent Dirichlet Allocation: A Local Differential Privacy Based Framework

Authors: Yansheng Wang, Yongxin Tong, Dingyuan Shi6283-6290

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

Reproducibility Variable Result LLM Response
Research Type Experimental Extensive experiments on three open datasets verified the effectiveness of our solution.
Researcher Affiliation Academia Yansheng Wang, Yongxin Tong, Dingyuan Shi SKLSDE Lab, BDBC, School of Computer Science and Engineering and IRI, Beihang University, China {arthur wang, yxtong, chnsdy}@buaa.edu.cn
Pseudocode Yes Algorithm. 1 shows the details of local sampling. Algorithm. 2 shows the details of global integration. Algorithm. 3 shows our RRP mechanism.
Open Source Code No The paper does not provide any statement or link indicating that open-source code for the methodology is available.
Open Datasets Yes We use three open datasets: Reviews 2, Emails 3 and Sentiments 4 (Maas et al. 2011). The dataset Emails contains 33,716 spam/non-spam emails with M = 150 and |V| = 3309. The dataset Sentiments has 50,000 highly polar movie reviews with positive/negative sentiments, with M = 150 and |V| = 22574.
Dataset Splits Yes We split training data and test data by 4 : 1 for logistic regression and train both data for 100 iterations with the same solver.
Hardware Specification No The paper does not provide specific hardware details (e.g., CPU, GPU models, memory) used for the experiments. It only mentions general terms like 'computer clusters'.
Software Dependencies No The paper does not specify software dependencies with version numbers (e.g., Python 3.x, PyTorch 1.x, specific library versions).
Experiment Setup Yes Parameter settings. We randomly sample 1K, 5K and 3K instances respectively from Reviews, Emails and Sentiments for evaluation. The default ϵ is 7.5 for all datasets and the default K is 20 for Reviews, 30 for Emails and 50 for Sentiments.