Lecturer: Benjamin Paaßen
Fields: Machine Learning
Content
Machine learning is concerned with automatically learning models (patterns, regularities, correlations) from known data which generalize to new data. To do so, it combines concepts from mathematics (esp. statistics, probability theory, linear algebra, and optimization), artificial intelligence, and computer science. This course will provide an introduction to machine learning for the un-initiated. While some math will be necessary, everything will be accompanied by pictures and examples to get the core intuition across 😊
In more detail, the course will have four sessions with the following topics:
- Session 1: Basic Concepts: What is Machine learning and how does it relate to Artificial Intelligence? What are types of ML? What does ‘learning’ mean in ML? We will also discuss the basic ingredients of an ML algorithm (loss function, model class, and optimization strategy), linear regression as an example for such an algorithm, underfitting, overfitting (and how to prevent it), how probabilities help us to make precise what ‘generalization’ means, and how to design a basic ML experiment.
- Session 2: Classic machine learning tasks and methods to solve them: The distance perspective on ML, Regression, Classification, Dimensionality Reduction, Clustering, with respective methods for each task; and decision trees/forests
- Session 3: Artificial neural networks and deep learning: How to build artificial neural networks from single neurons to present-day transformers
- Session 4: Reinforcement learning and ethics
Each session is accompanied by a (voluntary) programming exercise in Python. Exercise sheets (and slides) can be found here: https://bpaassen.gitlab.io/Teaching.html
Literature
- This is optional literature for people who want to dive in deeper after the course:
- Biehl, M. (2023). The Shallow and the Deep: A biased introduction to neural networks and old school machine learning. https://www.cs.rug.nl/~biehl/
- Barber, D. (2012). Bayesian Reasoning and Machine Learning. Cambridge University Press. Cambridge, UK. http://www.cs.ucl.ac.uk/staff/d.barber/brml/
- Goodfellow, I., Bengio, Y., and Courville, A. (2016). Deep Learning. MIT Press. Cambridge, MA, USA. https://www.deeplearningbook.org/
- Paaßen, B., Artelt, A., Hammer, B. (2019). Lecture Notes on Applied Optimization. Bielefeld University. https://pub.uni-bielefeld.de/record/2935200
- Shalev-Shwartz, S., and Ben-David, S. (2014). Understanding Machine Learning: From Theory to Algorithms. Cambride University Press. Cambridge, UK. https://www.cs.huji.ac.il/w~shais/UnderstandingMachineLearning/
Lecturer
Benjamin Paaßen received their doctoral degree in intelligent systems in 2019 from Bielefeld University on the topic of ‘Metric Learning for Structured Data’. Afterwards, they received a DFG research fellowship for a stay at The University of Sydney in Australia and Humboldt-University of Berlin. From 2021-2024, they were deputy head of the educational technology lab at the German Research Center for Artificial Intelligence (DFKI). Since April 2023, they are junior professor for knowledge representation and machine learning (KML, speak ‘camel’) at Bielefeld University. Their research foci are machine learning on structured data and artificial intelligence for education.
Affiliation: Bielefeld University
Homepage: https://bpaassen.gitlab.io/