IMPROVING COMPUTATIONAL EFFICIENCY IN FEDERATED LEARNING: APPLICATIONS IN MACHINE LEARNING AND DATA MINING

Abstract

Federated Learning (FL) has emerged as a promising distributed machine learning paradigm that enables multiple devices to collaboratively train models without sharing raw data, thereby preserving privacy and enhancing data security. Despite its advantages, the practical deployment of FL is constrained by several computational challenges, including high communication overhead, intensive local computation, energy consumption, memory limitations, and statistical heterogeneity caused by non-IID data distributions. These issues are particularly critical in resource-constrained environments such as Internet of Things (IoT) devices, edge computing systems, and TinyML platforms.This survey presents a comprehensive review of recent advancements in improving computational efficiency in federated learning, with a focus on machine learning and data mining applications. The study analyzes 60 research articles published between 2022 and 2026 and examines key optimization techniques, including model compression through pruning, quantization, and knowledge distillation, feature selection and dimensionality reduction methods, gradient sparsification, clustered federated learning, and privacy-preserving mechanisms such as differential privacy and homomorphic encryption. The reviewed literature demonstrates that these approaches can significantly reduce communication costs, accelerate model convergence, and lower energy consumption while maintaining acceptable predictive performance. In many cases, communication overhead is reduced by 30–70%, with only minor accuracy degradation compared to centralized learning models.The survey further highlights the effectiveness of hybrid approaches that integrate clustering, compression, and sparsification strategies to address heterogeneity and scalability challenges. However, issues related to robustness, privacy-utility trade-offs, and cross-domain generalization remain open research problems. The findings emphasize the importance of computational efficiency for enabling scalable, practical, and sustainable federated learning systems in privacy-sensitive and resource-limited environments.