Towards Reciprocal Learning Theory: Can We Generalize From Self-Selected Data?

Rodemann, Julian; Jansen, Christoph; Schollmeyer, Georg und Bailie, James : Towards Reciprocal Learning Theory: Can We Generalize From Self-Selected Data? 2nd Workshop on Learning Under Weakly Structured Information (LUWSI), Tübingen, 7. April 2025.

Abstract

eciprocal learning (as introduced at the 1st LuWSI Workshop) generalizes several learning paradigms, ranging from active learning over multi-armed bandits to self-training. These methods not only learn parameters from data, but also vice versa: They iteratively alter the training data as a function of previously learned parameters. In my talk at the 2nd LuWSI Workshop, I will address the elephant in the room: How well can these algorithms generalize from such self-selected samples?

• Item Type: Speech
• Faculties: Mathematics, Computer Science and Statistics > Statistics; Mathematics, Computer Science and Statistics > Statistics > Chairs/Working Groups > Foundations of Statistics and their Applications
• Subjects: 000 Computer science, information and general works > 000 Computer science, knowledge, and systems; 500 Science > 510 Mathematics