CAS ETH AMI: Applied Machine Learning & Information Processing

Dr Oylum Akkus Ispir (external page ETH Zurich)

This module introduces the essential concepts and tools of data science without requiring any technical or mathematical background. The main purpose is to provide participants the basic knowledge and intuition to use data and understand how it is used.

The mathematical topics for data science and machine learning will be covered as well. The participants explore the data landscape, understand key data science techniques, and learn how to apply them. The key topics of this module are the types of data, sources, and collection methods, data lifecycle, data-driven decision making, exploratory data analysis, experimental testing, regression models, and machine learning.

This module also showcases real-world ethical dilemmas, leadership challenges, and communication strategies specific to the data science field. Through interactive discussions, and practical exercises, participants will develop a comprehensive understanding of how to navigate ethical complexities, lead data-driven teams with integrity, and communicate technical information effectively to diverse audiences. 

Each topic will be enriched with collaborative discussions and hands-on exercise, enabling participants to develop a practical understanding of broad data science topics. 

Dr Oylum Akkus Ispir
Guney Tombak (Computer Vision Lab, ETH Zurich)

This module introduces participants to AI-assisted programming using modern chatbots and the emerging paradigm of autonomous agents. Participants learn how to translate real tasks into clear, testable prompts; use chatbots as “pair programmers” to write, debug, and refactor code; and build simple agent workflows that can plan multi-step actions, call tools, and validate results.

Through guided demonstrations and hands-on exercises (e.g., in Python/Colab), participants develop the confidence to use chatbots and agents productively in everyday programming and analytics work. This module also serves as a preparation for modeling and programming tasks in the other CAS modules.

Dr. Julio Jose Silva Rodríguez (Computer Vision Lab, ETH Zurich)

This module covers the essential concepts and tools of machine learning. It provides a rigorous, practice-oriented introduction to modern machine learning, combining core concepts with hands-on exercises.
Participants will understand how learning systems generalize from data, how to choose and evaluate models, and how to reason about performance under real-world constraints. The module covers the fundamentals of machine learning, approaches for low-data learning (including practical strategies when labeled data are limited), and the principles behind neural networks, supported by guided coding sessions. It also introduces key ideas in self-supervised learning and attention-based models, highlighting why these methods have become central to today’s foundation models and state-of-the-art applications.
Throughout the module, participants will complete guided Google Colab exercises to implement models, run experiments, interpret results, and develop practical confidence with modern ML workflows.

Professor Benjamin Grewe (external page Grewe Lab, ETH Zurich) 

This module introduces reinforcement learning (RL) as a principled framework for sequential decision-making in which an agent learns through interaction with an environment by optimizing long-term reward signals.
The module develops the foundations of RL from both a neuroscience and a machine-learning perspective: it begins with defining the links between reinforcement learning and learning in the brain, including canonical experimental paradigms and key update rules such as the Rescorla–Wagner and temporal-difference formulations, leading to the core intuition behind Q-learning.
It then transitions to reinforcement learning in machines, covering the fundamental components of RL systems and the formalism of policies and value functions, as well as the standard mathematical framework of Monte Carlo methods, Markov reward processes, and Markov decision processes, culminating in the Bellman (expectation) equation as the central tool for reasoning about optimal behavior.