New & Events
SUNRISE-6G proudly announces a significant milestone in Edge Computing and Machine Learning. The project’s first publication, “Unlocking the Path Towards Automation of Tiny Machine Learning for Edge Computing”, presents innovative advances in automated aspects of TinyML.
The publication was accepted at the 2024 IEEE International Conference on Smart Applications, Communications and Networking (SmartNets24), which took place from 28 to 30 May 2024, and presents the innovative advances made by the SUNRISE-6G team. At its core, the research explores the automation of Tiny Machine Learning (TinyML) through compression techniques. By compressing pre-trained ML models that can seamlessly fit into different target environments, such as edge devices, and be reliably deployed via the Pi-Edge platform, the project sets a new standard for efficiency and accessibility in edge computing.
A key highlight of our publication is the presentation of an IoT use case for anomaly detection, demonstrating the practical applicability of our developed methods. Our research includes a comprehensive evaluation of the accuracy of several ML models, including an innovative generative AI ground model with zero-shot inference functions. We meticulously analyse the trade-off between compression size and accuracy and provide insights into optimal deployment strategies for edge intelligence applications.
The paper focuses on automated Machine Learning (AutoML) and its application in Tiny Machine Learning, a subset of Artificial Intelligence tailored to resource-constrained devices. We explore the integration of AutoML techniques with emerging technologies such as Big Data, the Internet of Things (IoT) and Edge Computing, emphasising the critical role of Edge Intelligence in processing data closer to the source. This approach significantly reduces latency and enhances efficiency in distributed systems.
The SmartNets24 conference provided a platform to share these revolutionary findings. This year’s event was held at James Madison University in beautiful Harrisonburg, Virginia (US), which brought many bright minds together, especially from the US. You could also see how closely academia and industry work and research innovative aspects, such as smart cities, intelligent transportation, etc. Another positive outcome was that the publication was one of the main pillars of most talks and presentations since sustainable AI and eco-ML through TinyML were at the core of this year’s spirit. All this was done with the performance, reliability, and accuracy of AI in mind, which was also the direction of our publication.
The team looks forward to continued collaboration and realising a future where seamless connectivity and intelligent edge solutions empower communities worldwide.
For more information, please visit [link to the publication]