client-selection

Source code for the paper "Joint Class-Balanced Client Selection and Bandwidth Allocation for Cost-Efficient Federated Learning in Mobile Edge Computing Networks". This paper is pulished in IEEE TMC.

[ACM e-Energy'24] Code and data for "FedZero: Leveraging Renewable Excess Energy in Federated Learning"

Source code for the paper "Energy-Efficient Client Sampling for Federated Learning in Heterogeneous Mobile Edge Computing Networks", this paper is pulished in ICC 2024.

[KDD 2025] Proxy-Validated Importance-Aware Federated Sample Selection with Meta Learning

Random Client Selection FedAvg Federated Learning

A robust Federated Learning framework implementing the novel FedCADS-UCB algorithm. Engineered for resilient client selection using CUSUM drift detection, adaptive multi-armed bandits, and hierarchical clustering to maintain high accuracy (>95%) during concept drift and label poisoning attacks.

This artifact provides a complete and runnable implementation of a federated learning client selection system inspired by DCS (Accepted in IEEE PerCom.

This repository explores Federated Learning (FL) with a focus on FedAvg, client heterogeneity, and novel client selection strategies. We conduct experiments using CIFAR-100 and Shakespeare datasets with PyTorch.

Federated Learning with client selection strategies (Random, Loss-based, Power-of-Choice), domain generalisation and model pruning. Applied to image classification (FEMNIST) and semantic segmentation (IDDA). PyTorch.

A PyTorch framework for drift-resilient Federated Learning using Shapley values, UCB bandits, and CUSUM drift detection (FedCADS-UCB).

Energy-aware federated learning framework for heterogeneous IoT — joint computation & communication optimization achieving a 59% reduction in total energy vs. FedAvg baseline while maintaining ~91-92% accuracy (Flower + PyTorch, MNIST)

Code regarding the Semantic Segmentation in Federated Learning project for the Machine Learning and Deep Learning 2022/2023 project.