russelljeffrey/TPFL

TPFL

This repository is the implementation of TPFL: Tsetlin-Personalized Federated Learning with Confidence Based Clustering, a novel method that implements Tsetlin Machine (TM) algorithm in Personalized Federated Learning (PFL) context for the first time. TPFL is implemented under 5 experimental setups. Paper >>> https://www.arxiv.org/abs/2409.10392

Jupyter Notebook files:

• TPFL_Experiments.ipynb: TPFL Implementation
• FLIS_Baseline.ipynb: FLIS Implementation
• FedAvg_FedProx_Baselines.ipynb: FedAvg and FedProx Implementation
• IFCA_Baseline.ipynb: IFCA Implementation
• Model_Size_Report.ipynb: Model Communication Cost Implementation

baselines used for comparison with TPFL:

1. FedAvg
2. FedProx
3. IFCA
4. FLIS
5. FedTM

Usage

Run TPFL_Experiments.ipynb notebook. We ran the notebooks on Google Colab Pro.

Results

To see the results, please go the TPFL Report directory.