Thema der Dissertation:
Efficient Operation of Edge AI and Tiny Machine Learning on Heterogeneous Resource-Constrained Devices Thema der Disputation:
Bringing Artificial Intelligence to Low-power Hardware: Challenges & Trends
Efficient Operation of Edge AI and Tiny Machine Learning on Heterogeneous Resource-Constrained Devices Thema der Disputation:
Bringing Artificial Intelligence to Low-power Hardware: Challenges & Trends
Abstract: Tiny Machine Learning (TinyML) brings deep learning to resource-constrained, microcontroller-based devices, reducing communication overhead and enabling responsive, privacy-preserving Artificial Intelligence of Things (AIoT) applications. Deploying and maintaining deep learning models on these devices, however, remains challenging due to tight resource budgets and fragmented hardware and software ecosystems. In this talk, I will introduce Tiny Machine Learning Operations (TinyMLOps), the domain that extends machine learning lifecycle management to resource-constrained devices. I will provide an overview of research gaps and challenges in this field and give notions of how different model optimization techniques and frameworks for model deployment and execution/inference mitigate these challenges. The talk connects model-level optimization techniques with system-level deployment considerations, highlighting their trade-offs in predictive performance, resource usage, and inference latency.
Zeit & Ort
05.10.2026 | 13:00
Seminarraum 005
(Fachbereich Mathematik und Informatik, Takustr. 9, 14195 Berlin)
&
WebEx