Aktuelles

Beschreibung

We will focus on what can help to build learning agents interacting with an initially unknown world. Starting with asymptotically optimal yet infeasible methods showing the ultimate limits of machine learning and artificial intelligence, we'll work all the way down to practically useful tricks. Since the world is dynamic, unlike many other machine learning courses ours will put strong emphasis on learning to deal with sequential data: we do not just want to learn reactive input / output mappings but programs (running, e.g., on recurrent neural nets) that perceive, classify, plan, make decisions, etc.

Weitere Informationen in der ML & BIO II WWW-Seite. Dort finden Sie auch weiterführende Links zu folgenden Themen: Robot learning at the TUM Cogbotlab of TUM Informatics, recurrent neural networks (RNNs), optimal ordered problem solver, optimal universal learners, Goedel machines and other self-referential meta-learners, reinforcement learning, artificial evolution, evolving RNNs for robot control, genetic programming, non-linear ICA, artificial curiosity, non-halting Turing machines and generalizations of Kolmogorov complexity, a complexity-based theory of beauty, low-complexity art (a new minimal art form based on algorithmic information theory), the speed prior for optimal computable inductive inference in quickly computable universes, an algorithmic theory of everything, probabilistic robotics, resilient robots, robot cars, learning attentive vision, hierarchical learning, time series prediction, financial forecasting, artificial music composition, artificial ants, the New AI: general & sound & relevant for physics. Vergleiche Publikationen und deutsche Seite. Konvergiert die Menschheitsgeschichte?

Material

Literaturhinweise