Written exam (in case of a small number of participants: oral tests)

This lecture builds on the available knowledge how animals and humans plan,
decide on, and control their behavior and how they progressively optimize and
adapt their behavior over time. Accordingly, algorithms are introduced for behavioral
decision making, control, optimization, and adaptation. In particular,
the lecture introduces spatial representations for behavioral control, forwardinverse
control models, including the learning of such representations and models.
Also the encoding and the learning of motor control primitives and motor
complexes is considered. Last but not least, self-motivated artificial systems
are considered that strive to maintain internal homeostasis and to maximize
information gain.

Students know how intelligent behavior can be generated and learned in artificial
systems. They can apply reinforcement learning (RL), including hierarchical
RL, factored RL, and actor-critic approaches to the appropriate problems. Moreover,
they are aware of the contrast between model-free and model-based RL
approaches. They know about dynamic motion primitives and know how to optimize
them. Moreover, they know about Gaussian Mixture Models, including
how to learn and optimize them. They can implement information-gain driven
and self-motivated behavior and are aware of the exploration-exploitation
dilemma. Moreover, they are aware of model-predictive control, of options to
learn suitable model-predictive structures, and of options to suitably abstract
such structures. Finally, they know how sensorimotor-grounded spatiotemporal
representations can be learned, stored as episodic memory units, and can be
abstracted into cognitive maps, enabling model-based RL.

Literatur / Literature:
Wird in der Veranstaltung ausgegeben (Buchkapitel und Artikel in Englisch). / Will be supplied (book chapters and papers in English).

Voraussetzungen / Prerequisites:
Vorkenntnisse in maschinellem Lernen, Künstlichen Neuronalen Netzen, Deep Learning oder Künstlicher Intelligenz sind notwenig. / Knowledge about machine learning, artificial neural networks, deep learning, or artificial intelligence is required.