Machine Unlearning of Federated Clusters

Authors: Chao Pan, Jin Sima, Saurav Prakash, Vishal Rana, Olgica Milenkovic

ICLR 2023 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental 6 EXPERIMENTAL RESULTS
Researcher Affiliation Academia Department of Electrical and Computer Engineering University of Illinois Urbana-Champaign, USA
Pseudocode Yes Algorithm 1 Secure Federated Clustering
Open Source Code Yes Our implementation for the proposed method is available at https://github.com/thupchnsky/mufc.
Open Datasets Yes The seven datasets used in our simulations are all publicly available. Among these datasets, TCGA and TMI contain potentially sensitive biological data and are downloaded after logging into the database. We adhered to all regulations when handling this anonymized data and will only release the data processing pipeline and data that is unrestricted at TCGA and TMI. Datasets that do not contain sensitive information can be downloaded directly from their open-source repositories.
Dataset Splits No The paper does not provide specific dataset split information (exact percentages, sample counts, citations to predefined splits, or detailed splitting methodology) needed to reproduce the data partitioning for training, validation, and testing.
Hardware Specification No The paper does not provide specific hardware details (exact GPU/CPU models, processor types with speeds, memory amounts, or detailed computer specifications) used for running its experiments.
Software Dependencies No The paper does not provide specific ancillary software details (e.g., library or solver names with version numbers) needed to replicate the experiment.
Experiment Setup Yes The parameter γ > 0 determines the number of quantization bins in each dimension. If the client data is not confined to the unit hypercube centered at the origin, we scale the data to meet this requirement. Then the number of quantization bins in each dimension equals B = γ 1, while the total number of quantization bins for d dimensions is Bd = γ d. The symbol K in Fig. 3 represents the maximum number of (true) clusters among clients, while K represents the number of true clusters in the global dataset. ... During training, the clustering parameter K is set to be the same in both clients and server for all methods.