Tiny Federated Learning for Constrained Sensors: A Systematic Literature Review
Abstract
The recent tiny federated learning (TinyFL) paradigm has generated significant interest in enabling resource-constrained edge devices to perform local machine learning training. TinyFL reduces latency, communication costs, and privacy risks by sharing model weights or gradients instead of raw data. Our research provides a systematic literature review (SLR) of 30 selected studies to understand the current research landscape in TinyFL. The review highlights that most studies focus on Internet of Things (IoT) environments, using lightweight architectures and optimization techniques to overcome challenges related to limited memory, processing power, and energy resources. While most studies address generic IoT applications, emerging efforts in domains, such as the Internet of Medical Things and unmanned aerial vehicles, highlight the adaptability of TinyFL to diverse scenarios. Common strategies include using FedAvg to aggregate learning models, such as multilayer perceptron, convolutional neural networks, and deep neural networks. Frameworks, such as TensorFlow Lite Micro and Python, enable local training on a diverse range of devices, including microcontrollers, such as the ESP32 and Arduino Nano BLE 33, as well as more powerful single board, such as the Raspberry Pi 4 Model B. The limitation of data storage on devices and the security risks inherent in sharing models are gaps identified in the analysis of the studies. In addition to providing a current overview of existing…
Read the paper
Related studies
- Active Learning during Federated Learning for Object Detection
- Fairness-aware federated learning
- Decentralized Federated Learning
- Building a Personally Identifiable Information Recognizer in a Privacy Preserved Manner Using Automated Annotation and Federated Learning
- Preservation of Patient Level Privacy: Federated Classification and Calibration Models
- Efficient explainable federated learning
Go deeper in Evidence Hub
Evidence Hub adds a study data sheet (design, sample size, evidence level, key numbers), an AI summary and a chat grounded in the paper's text.