Notice: The reproducibility variables underlying each score are classified using an automated LLM-based pipeline, validated against a manually labeled dataset. LLM-based classification introduces uncertainty and potential bias; scores should be interpreted as estimates. Full accuracy metrics and methodology are described in [1].

Personalized Dictionary Learning for Heterogeneous Datasets

Authors: Geyu Liang, Naichen Shi, Raed AL Kontar, Salar Fattahi

NeurIPS 2023 | Venue PDF | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental In this section, we showcase the effectiveness of Algorithm 1 using synthetic and real data. All experiments are performed on a Mac Book Pro 2021 with the Apple M1 Pro chip and 16GB unified memory for a serial implementation in MATLAB 2022a.
Researcher Affiliation Academia Geyu Liang University of Michigan Ann Arbor, MI 48109 EMAIL Naichen Shi University of Michigan Ann Arbor, MI 48109 EMAIL Raed Al Kontar University of Michigan Ann Arbor, MI 48109 EMAIL Salar Fattahi University of Michigan Ann Arbor, MI 48109 EMAIL
Pseudocode Yes Algorithm 1 Per MA: Federated Matching and Averaging
Open Source Code No The paper does not provide a direct link to open-source code for the described methodology or state that the code is available in supplementary materials.
Open Datasets Yes In section 5.2, we aim to learn a dictionary with imbalanced data collected from MNIST dataset (Le Cun et al., 2010).
Dataset Splits No The paper describes the construction of synthetic and MNIST datasets but does not explicitly provide training/test/validation dataset splits or percentages for reproducibility.
Hardware Specification Yes All experiments are performed on a Mac Book Pro 2021 with the Apple M1 Pro chip and 16GB unified memory for a serial implementation in MATLAB 2022a.
Software Dependencies Yes All experiments are performed on a Mac Book Pro 2021 with the Apple M1 Pro chip and 16GB unified memory for a serial implementation in MATLAB 2022a.
Experiment Setup Yes In this section, we validate our theoretical results on a synthetic dataset. We consider ten clients, each with a dataset generated according to the model 1. The details of our construction are presented in the appendix. Specifically, we use ΞΆ = 0.15 for the experiments in Section 5.1.