Optimizing Edge AI Models with High Performance Computing
Getting small by going big
Learn how AI models can be tailored to run efficiently on small, embedded devices.
Limited in size, embedded devices offer numerous benefits such as reduced latency, lower costs, enhanced reliability, and independence from cloud providers. The main challenge lies in reducing the model's size to fit these compact devices without significantly compromising its accuracy. Through this course, you'll learn how high-performance computing (HPC) can be utilized to effectively shrink your AI models. We'll focus on using parallel hyperparameter optimization—a method that helps determine the best settings for AI models—to ensure they perform well even within the constraints of smaller hardware.
This event is part of the European funded RebootSkills project that aims to facilitate access for the manufacturing industry to high-class training in digital skills.
Your trainer: Kurt De Grave
Dr. ir. Kurt De Grave is a senior researcher in artificial intelligence. He holds a Ph.D. in Computer Science and a Master in Engineering Science, both from KU Leuven. He co-invented several machine learning algorithms including the Neighborhood Subgraph Pairwise Distance Kernel (NSPDK), kLog (a declarative environment for machine learning in multi-relational data), the prediction of odor pleasantness from chemical structure (Science, 2017), and robot scientist Eve. Eve’s most significant discovery is that triclosan is an inhibitor of wild-type and drug-resistant dihydrofolate reductase in the malaria-causing parasites P. falciparum and P. vixax. Eve’s approach for closed-loop optimisation has been widely copied by the pharmaceutical industry and in materials science. In December 2015, Kurt joined Flanders Make, the strategic research centre for the manufacturing industry. There, he designs artificial intelligence algorithms for industrial innovations such as the optimization of metal sheet shaping, product inspection, and additive manufacturing processes. Kurt has been using HPC clusters for his research since 2007.
Practical Information
- Tuesday November 19th
- 10-12 am, Online (Leuven, CEST) | Convert to your time zone
- Language: English
- Participation is free of charge, registration mandatory via the form at the bottom of this page
Fill out this form to rewatch the training
