Lecture: Machine Learning

Overview

This course introduces students to the fundamental concepts, techniques, and algorithms in machine learning. It covers the mathematical and theoretical foundations, supervised and un-supervised learning techniques, evaluation methods, and advanced aspects. Students will gain hands-on experience in implementing, training, and optimizing machine learning models using real-world datasets.

Organization

Lecturer: Prof. Dr. A. Bojchevski
Time: Tuesdays, 16:00 - 17:30  and  Wednesdays, 16:00 - 17:30
Place: Hörsaal II, Physics Institute

Seminar: Trustworthy Machine Learning

Machine learning models are increasingly used in safety-critical applications and to make automated decisions about humans. Beyond accuracy and efficiency, we expect such models to also be robust to noise and adversaries, to faithfully represent their (aleatoric and epistemic) uncertainty, to preserve privacy, to be fair w.r.t. different demographic groups,  and to be interpretable. In this seminar, we will cover the latest research on these trustworthiness aspects, as well as the (fundamental) trade-offs between them. We will study the shortcomings and failures of traditional machine learning models and how to improve them.

Organization

Lecturer: Prof. Dr. A. Bojchevski
Time: Mondays, 08:00 - 09:30
Place: Room 1.421, Building 415 (Sibille-Hartmann-Str. 2-8)
 

Lecture: Advanced Machine Learning

Overview

In the Advanced Machine Learning lecture we will cover various advanced concepts, techniques, and algorithms. We will place focus both on the mathematical and theoretical aspects, as well as the practical aspects which involve implementing, training, and optimizing machine learning models using real-world datasets. The lecture is organized in four parts. In the first part we will tackle machine learning for graph data including generative models, ranking, and graph neural networks. In the second part we will cover modern generative models such as variational autoencoders, normalizing flows and generative adversarial networks. In the third part we will cover robustness including both attacks on machine learning models (evasion, poisoning) and defenses (certificates). Finally, in the last part we will cover uncertainty quantification tech- niques such as Bayesian neural networks, Gaussian processes and conformal prediction. Solid background in the fundamentals of machine learning is highly recommended, e.g. you should have passed our “Machine Learning“ lecture or equivalent.
Literatur

1. “Probabilistic Machine Learning: An Introduction“ von Kevin Patrick Murphy
2. “Probabilistic Machine Learning: Advanced Topics“ von Kevin Patrick Murphy

Organization

Lecturer: Prof. Dr. A. Bojchevski
Place: Hörsaal II, Physics Institute

Seminar: Adversarial Machine Learning

Overview

In the Adversarial Machine Learning seminar, we will explore the robustness of machine learning models. This seminar will have a practical focus where the students will be split into two sets of teams. One set of teams will focus on developing various attacks to break or manipulate machine learning models, e.g. given an image of a cat design algorithms to add impercetible adversarial noise to the input to cause the model to misclassify it as a dog. The other set of teams will focus on defending against such attacks. The attacks and defenses will be carried out in multiple rounds allowing the attackers and defenders to learn from each other to improve their approach.

Organization

Lecturer: Prof. Dr. A. Bojchevski
Place: Room 1.421, Building 415 (Sibille-Hartmann-Str. 2-8)
Introductory Presentation:  Slides

Lecture: Machine Learning

Overview

This course introduces students to the fundamental concepts, techniques, and algorithms in machine learning. It covers the mathematical and theoretical foundations, supervised and un-supervised learning techniques, evaluation methods, and advanced aspects. Students will gain hands-on experience in implementing, training, and optimizing machine learning models using real-world datasets.

Organization

Lecturer: Prof. Dr. A. Bojchevski
Place: Hörsaal II, Physics Institute

Seminar: Limitations of Large Language Models

This seminar explores the critical limitations of Large Language Models (LLMs) through the study of: 
Jailbreaking: How LLMs can be intentionally manipulated to bypass safeguards and restrictions, leading to unintended or unethical outputs. Hallucinations: The tendency of LLMs to generate confidently incorrect or fabricated information, undermining their reliability. Reasoning: Gaps in logical coherence and contextual understanding that affect the models' ability to perform consistent and accurate reasoning. Scalability: Challenges related to the increasing computational and environmental costs of training larger models, and the diminishing returns on performance improvements. We will also examine other aspects that underscore the limitations of LLMs, providing a comprehensive perspective on their current capabilities and future directions.

Organization

Lecturer: Prof. Dr. A. Bojchevski
Place: Room 1.421, Building 415 (Sibille-Hartmann-Str. 2-8)
 

Lecture: Advanced Machine Learning

Overview

In the Advanced Machine Learning lecture we will cover various advanced concepts, techniques, and algorithms. We will place focus both on the mathematical and theoretical aspects, as well as the practical aspects which involve implementing, training, and optimizing machine learning models using real-world datasets. The lecture is organized in four parts. In the first part we will tackle machine learning for graph data including generative models, ranking, and graph neural networks. In the second part we will cover modern generative models such as variational autoencoders, normalizing flows and generative adversarial networks. In the third part we will cover robustness including both attacks on machine learning models (evasion, poisoning) and defenses (certificates). Finally, in the last part we will cover uncertainty quantification tech- niques such as Bayesian neural networks, Gaussian processes and conformal prediction. Solid background in the fundamentals of machine learning is highly recommended, e.g. you should have passed our “Machine Learning“ lecture or equivalent.
Literatur

1. “Probabilistic Machine Learning: An Introduction“ von Kevin Patrick Murphy
2. “Probabilistic Machine Learning: Advanced Topics“ von Kevin Patrick Murphy

Organization

Lecturer: Prof. Dr. A. Bojchevski
Place: Hörsaal II, Physics Institute

Seminar: AI Security

Overview

This seminar explores the rapidly evolving field of AI Safety and the emerging challenges. The topics include designing robust and reliable AI systems, adversarial threats, certification and verification, alignment (e.g. with human values), explainability, monitoring and preventing accidents or misuse, agentic and multimodal safety, as well as AI and society. The goal is to give participants broad insight into AI safety for a well-rounded understanding of the field. We will study the latest research via student-led presentations on papers and topics of their choice and panel discussions.

Organization

Lecturer: Prof. Dr. A. Bojchevski
Place: Room 1.421, Building 415 (Sibille-Hartmann-Str. 2-8)