Embedded Machine Learning
Introduction
Machine learning (ML) enables us to teach computers the ability to learn from experience E with respect to some tasks T and performance measure P, without being explicitly programmed. Machine learning has been proved that it can be very effective to make sense of sensor data, especially with deep neural networks.
The rise of the Internet of Things (IoT) shifts the future of machine learning from clouds toward the edge with small, embedded devices. There is a variety of reasons for this including communication cost, coverage, latency, privacy, security and safety.
While machine learning usually requires heavy-duty power-hungry processors and a large dataset, embedded devices have limited computing power and small dataset. Making machine learning algorithms efficient to run on resource-constrained devices like microcontrollers for a year even more with a single coin battery is very demanding, especially for real-time applications with image and speech processing such as image recognition, segmentation, object localisation, multi-channel speech enhancement, and speech recognition.
Suggested Topics
Topics of particular interest include, but are not limited to:
Machine learning for embedded devices
Compression of neural networks for inference deployment, including methods for quantization (including binarization), pruning, knowledge distillation, structural efficiency and neural architecture search
Learning on edge devices, including federated and continuous learning
Learning from limited labelled data, including methods for self-taught learning, transfer learning, unsupervised learning, generative models, and latent representation learning
Exploring new ML models designed to use on designated device hardware
New benchmarks suited to edge devices and learning on the edge scenarios
New and emerging applications that require the use of ML on resource-constrained hardware
Embedded systems for machine learning
Future emerging processors and technologies for use in resource-constrained environments
Hardware security/privacy
Low-power wireless systems
Energy management and smart grids
Network on a chip
Further reading
Survey paper: Deep Learning on Mobile and Embedded Devices: State-of-the-art, Challenges, and Future Directions
Some possibly interesting assignments can be found on the Pervasive Systems Education page . Examples includes
Luisterving 1 - Activity recognition in Nature using Sound and AI
Luisterving 2 - Activity recognition in Nature using Sound and AI
Distributed Wireless Learning: Edge Ai for a Better Wireless Network
Holisitc Learning for Wireless One Ai to Rule Them All (The Networks)
Compressed Convolutional Neural Networks for The Classification Of Electrocardiogram Signals on Embedded Devices
Weighted Convolutional Neural Networks Rare Electrocardiogram Signals Detection for Real-Time Heart Monitoring
Smartphone-Based Indoor Localization using Compressive Transformer
Compressed Generative Adversarial Networks for Real-Time Feature Extraction for Fingerprinting Localization with Smartphones
Convolutional Algorithms for Real-Time Data Processing with Time Series on Smartphones
Lightweight Image-Based Key Point Detection for Real-Time Bridge Monitoring with Smartphones
Embedded Machine Learning for Real-Time Flow Rate Measurement
Algorithms for Electric Vehicle Flexibility Aggregation for Optimal Charging
Data driven clustering of households in the electricity grid to support the energy transition
Information
For further information on the content of this track, you may contact the track chair: Le Viet Duc.