Optimized Artificial Intelligence in Edge Computing and IoT - January 2025

Length:  10 hours - 2 cfu

Abstract:

Artificial intelligence is employed across various scientific and engineering domains, from human-machine interaction to autonomous robotics, achieving human-level performance in many complex tasks. In industrial settings, AI optimizes production lines, predicts machinery failures, and develops smart services. However, adopting these solutions is challenging due to performance, resource, and privacy requirements. For example, low-precision logic supports deploying neural networks on tiny processors but could limit explainable AI (XAI), restricting edge processors in critical applications like autonomous driving, healthcare, and smart monitoring.

This course will cover techniques for developing efficient solutions on resource-constrained hardware through lectures, discussions, and practical implementations. Students will gain essential knowledge and skills to build explainable and adaptable AI systems. As an advanced course, basic knowledge of neural networks and optimization methods is preferable, but key deep learning concepts and standard metrics will be summarized initially.

Main topics:

      • Explainable AI (XAI): methods and approaches.

      • Efficient information representation of deep neural networks for edge devices.

      • Optimization techniques for AI models on resource-constrained hardware.

      • State-of-the-art generative modeling techniques for industrial and environmental scenarios.

      • Advanced techniques for training with low data, e.g., self-supervised and unsupervised pre-training.

      • Current research trends in advanced AI.

Dates & Venue

Suggested Readings:

Lecturer:

Dr. Pasquale Coscia - Dipartimento di Informatica

Assessor:

Dr. Pasquale Coscia - Dipartimento di Informatica