Modulcode:	infEDAI-01a
Englische Bezeichnung:	Tiny Machine Learning (Edge AI)
Modulverantwortliche(r):	Prof. Dr. Olaf Landsiedel
Turnus:	unregelmäßig (WS23/24 WS24/25)
Präsenzzeiten:	3V 2Ü 1PÜ
ECTS:	8
Workload:	45 h lectures, 30 h exercises, 15 h practical exercises, 150 h self studies
Dauer:	ein Semester
Modulkategorien:	BSc-Inf-WP (BSc Inf (21)) MSc-Inf-WP (MSc Inf (21)) 2F-MEd-Inf-WP (MEd-Hdl Inf (21)) 2F-MA-Inf-WP (2F-MA Inf (21)) MSc-WInf-WP-Inf (MSc WInf (21))
Lehrsprache:	Englisch
Voraussetzungen:	infCN-01a infOS-01a infDaSci-01a

Kurzfassung:

This course is a course at the intersection of Machine Learning and Embedded IoT Devices. The pervasiveness of ultra-low-power embedded devices, coupled with the introduction of embedded machine learning frameworks, will enable the mass proliferation of AI-powered IoT devices. The rapid growth of machine learning and how easy it is to use machine learning platforms make it an important subject for computer science students to learn about today.

Lernziele:

At the end of the course, students will have learnt the following:

• Practical experience through hands-on project assignments
• Gained familiarity with cutting-edge literature and frameworks in the field of TinyML
• Know how to train and deploy models on microcontrollers
• Conceived and developed a (novel) TinyML application running on a MCU

Lehrinhalte:

This course is a deep dive into efficient machine learning techniques that enable powerful deep learning applications on resource-constrained devices.

Topics cover

• Brief recap / introduction to ML and IoT
• efficient inference techniques, including model compression, pruning, quantization, neural architecture search, and distillation;
• and efficient training techniques, including distributed training, gradient compression and on-device transfer learning;
• followed by application-specific model optimization techniques for video, point cloud, generative model, NLP and LLM;

Our lectures provide students with the required fundamentals, and exercise and projects give students a hands-on experience in developing embedded AI systems & applications and exploring their real-world challenges. This course offers learning experiences that involve hands-on experimentation and analysis as they reinforce student understanding of concepts and their application to real-world problems. Overall, this course provides the students the ability to understand fundamental issues in the design of methods for embedded AI systems and applications.

Weitere Voraussetzungen:

A student should have taken courses on Internet of Things or embedded systems, such as the course "Internet Things and Wireless Networks" at CAU. Further, we expect students to have taken a course on machine learning and especially neural networks (including knowledge of backpropagation and experience in TensorFlow, PyTorch or similar platforms). Moreover, C programming skills are expected, as we will be using embedded IoT devices.

Prüfungsleistung:

The grade is determined by a final written exam. If the number of students registered for the course is less than 10, we might switch to an oral examination instead. A switch will be announced within the first two weeks of the course.

Prerequisite for admission to the exam: received at least 50% of homework assignment points and 50% of the project points. Also, the exercises might include a Prelab that has to be passed before the actual homework exercises can be conducted. Exceptions, for example, due to sickness, have to requested in due time before the exams. Exam admissions stay valid also for future terms. Individual, passed homework series apply only for the current term and are not transferable to future terms.

The final grade for the module is given by either 1) the exam grade or 2) 80% of the exam grade + 20% of the homework and project grade, whichever is the better of the two (as long as the exam is passed).

Lehr- und Lernmethoden:

Lectures, weekly exercises with software development, and a project.

Verwendbarkeit:

Literatur:

TBA in the course

Verweise:

https://www.ds.informatik.uni-kiel.de/en/teaching/tinyml-edge-ai/tinyml-edge-ai

Kommentar: