Foundations of Machine Learning

The "Foundations of Machine Learning" research area at ELIZA delves into the theoretical underpinnings and mathematical principles that drive the development of machine learning algorithms and systems. This area crucially advances the field of Artificial Intelligence by establishing a robust scientific basis for how machines learn from data. It encompasses a wide range of topics aimed at understanding and improving the fundamental capabilities of learning algorithms. Notably, this area includes ML-driven disciplines such as computer vision, natural language processing and robot learning, thereby broadening the theoretical focus to more applied domains within AI. PhD students associated with ELIZA are actively engaged in this research area, contributing to a deeper comprehension of learning processes and the development of novel theoretical frameworks.

Key aspects explored within the Foundations of Machine Learning at ELIZA include:

The development of new learning algorithms and techniques, which involves creating innovative approaches for machines to learn from data, and ultimately more efficient, accurate, and robust AI systems.
Theoretical analysis of learning algorithms, to better grasp the mathematical properties of existing and new algorithms, such as their convergence rates, generalization abilities, and computational complexity.
Understanding the principles of generalization, wherein the central question is why and when machine learning models are able to perform well on unseen data. Central to this aspect are concepts like statistical learning theory and VC dimension.
Addressing fundamental challenges in machine learning, such as limited data, noisy or biased datasets, and the interpretability and explainability of machine learning models.
Exploring different learning paradigms like online learning, reinforcement learning, probabilistic machine learning, and learning with different types of data or supervision.
Bridging the gap between neural networks and symbolic AI through the interdisciplinary area of neural-symbolic computing.

ELIZA scholars working in this area interact with the broader scientific community through publications in top-tier journals and presentations at prestigious conferences such as NeurIPS (Neural Information Processing Systems), ICML (International Conference on Machine Learning), COLT (Conference on Learning Theory), and AISTATS (Artificial Intelligence and Statistics). The investigations conducted within the Foundations of Machine Learning at ELIZA are critical for building more reliable, efficient, and intelligent AI systems, and for positioning Germany at the forefront of AI innovation. This research area provides the essential theoretical groundwork that underpins advancements in various AI applications.