Learning Model-Based Privacy Protection under Budget Constraints

Authors: Junyuan Hong, Haotao Wang, Zhangyang Wang, Jiayu Zhou7702-7710

AAAI 2021 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental Our empirical studies on both synthetic and real datasets demonstrate that the proposed algorithm can achieve a superior utility with a given privacy constraint and generalize well to new private datasets distributed differently as compared to the hand-designed competitors.
Researcher Affiliation Academia 1 Department of Computer Science and Engineering Michigan State University, East Lansing, MI 48823, USA 2 Department of Electrical and Computer Engineering University of Texas at Austin, Austin TX 78712, USA
Pseudocode Yes Algorithm 1 Protected Stochastic Gradient Descent
Open Source Code No The paper does not include an unambiguous statement that the authors are releasing the code for the work described in this paper, nor does it provide a direct link to a source-code repository for their methodology.
Open Datasets Yes IPUMS. IPMUS-International census database (Ruggles et al. 2018) and its processed version is from (Lee and Kifer 2018). ... MNIST35. The MNIST dataset (Lecun et al. 1998) includes 70, 000 gray-scale handwritten digits.
Dataset Splits No We randomly select 20% (80%) of the data for testing (training). ... We use the official training-testing splits for evaluating the learning algorithms". While training and testing splits are mentioned, specific details for a separate validation split (percentages, counts, or explicit methodology for how *their* data was partitioned for validation) are not provided.
Hardware Specification No The paper does not provide specific hardware details such as exact GPU/CPU models, processor types, or memory amounts used for running its experiments.
Software Dependencies No The paper does not provide specific ancillary software details with version numbers (e.g., library or solver names with version numbers like Python 3.8, CPLEX 12.4) needed to replicate the experiment.
Experiment Setup Yes Unless otherwise specified, 20 units of hidden variables are used in each LSTM and ct CDP is the privacy measurement. ... The protector is trained by a total of 50 batches of 20 steps and 100 epochs (1 epoch includes 1 scheduler update and 5 projector updates).