Sebastian Gottwald

Post-doctoral researcher at the Institute of Neural Information Processing at Ulm University with a background in mathematics and physics.

Current research interests

Uncertainty in decision-making systems and algorithms.
Abstraction in artificial systems.
Autonomous agents in arbitrary environments.
Latent state models, attention, and memory attractor dynamics.

Current teaching

Berechenbarkeit und Komplexität: Vorlesungsnotizen
Learning Systems II: Lecture notes and simulations
Foundations and Concepts of Cognitive Systems Modeling
Bachelor and Master Theses: At the Institute of Neural Information Processing at Ulm University, we are offering Bachelor and Master projects on Machine Learning, Information Theory, and Decision Theory. Simply approach one of our group members if you are interested.

Shared projects and research notes

Two-kinds-of-free-energy--notebooks Supporting information for the article Gottwald, Braun. The two kinds of free energy and the Bayesian revolution (2020).
RDFC: Rate-distortion with fixed cardinality. An implementation and comparison of RDFC (Banerjee et al.) with standard rate-distortion algorithms.
Rate-Distortion. A comparison of standard algorithms to determine Shannon's rate-distortion trade-off (Blahut-Arimoto, gradient descent, mapping approach).
Tensorflow 2 examples. Simple tensorflow 2 implementations for function optimization and neural network training using different APIs (models.Sequential, tf.keras.layers, tf.keras.metrics, custom training).
CNNs with tensorflow. Implementation of a CNN from scratch using tensorflow 1.
pr_func. Python class to simplify calculus with discrete probability distributions and other functions (see also these examples).
BA_class. Python class that contains a collection of Blahut-Arimoto type algorithms
BMA. A modular implementation of Blahut-Arimoto algorithms for feed-forward systems of bounded-rational agents.
Coarse-graining and Free Energy. Some coarse-graining properties of free energy expressions.
Legendre transform and rate-distortion. The derivative of the rate-distortion curve is NOT the temperature.

Publications

Hack P., Lindig-Leon C., Gottwald S., Braun D.A. Thermodynamic fluctuation theorems govern human sensorimotor learning, 2023, Scientific Reports 13, 869 (2023) (arXiv)
Hack P., Braun D.A., Gottwald S. The classification of preordered spaces in terms of monotones: complexity and optimization, 2023, Theory and Decision, 94, 693–720 (2023) (arXiv)
Hack P., Gottwald S., Braun D.A. Jarzyski’s Equality and Crooks’ Fluctuation Theorem for General Markov Chains with Application to Decision-Making Systems, 2022, Entropy 2022, 24(12), 1731 (arXiv)
Hack P., Braun D.A., Gottwald S. Representing preorders with injective monotones, 2022, Theory and Decision, 93, 663–690 (2022) (arXiv)
Gottwald S., Braun D.A. The two kinds of free energy and the Bayesian revolution, 2020, PLOS Computational Biology, 16(12), Public Library of Science (arXiv)
Lindig-León C., Gottwald S., Braun D.A., Analyzing Abstraction and Hierarchical Decision-Making in Absolute Identification by Information-Theoretic Bounded Rationality, 2019, Frontiers in Neuroscience 13, www.frontiersin.org
Hihn H., Gottwald S., Braun D.A., An Information-theoretic On-line Learning Principle for Specialization in Hierarchical Decision-Making Systems, 2019, IEEE 58th Conference on Decision and Control, IEEE
Gottwald S., Braun D.A., Bounded rational decision-making from elementary computations that reduce uncertainty, 2019, Entropy 21, 4, 375, MDPI (arXiv)
Gottwald S., Braun D.A., Systems of bounded rational agents with information-theoretic constraints, 2019, Neural Computation 31, 2, p.440-476, MIT Press (arXiv)
Schach S., Gottwald S., Braun D.A., Quantifying Motor Task Performance by Bounded Rational Decision Theory, 2018, Frontiers in Neuroscience 12, www.frontiersin.org
Hihn H., Gottwald S., Braun D.A., Bounded Rational Decision-Making with Adaptive Neural Network Priors, 2018, Artificial Neural Networks in Pattern Recognition, Springer (arXiv)
Gottwald S., Two-term spectral asymptotics for the Dirichlet pseudo-relativistic kinetic energy operator on a bounded domain, 2018, Annales Henri Poincaré 19(12), Springer (arXiv)

Academic positions

Post-Doctoral Researcher, 2016 - present
Institute of Neural Information Processing, Ulm University
Doing research in information theory and machine learning, supervising an interdisciplinary team of PhD students, giving lectures and exercise sessions.
Teaching Assistant, 2013 - 2016
Faculty of Mathematics, Informatics and Statistics, LMU Munich
Organizing exercise and tutorial groups, creating exams and exercise sheets, hiring and supervising tutors, and substituting lecturers for various lectures (see teaching experience).
Student Assistant, 2007 - 2013
Department of Mathematics and Faculty of Physics, LMU Munich
Tutoring lower semester students; correcting exams and exercise sheets; writing lecture notes.

Education

Dr.rer.nat (Mathematics, LMU Munich), 2013 - 2016
Thesis: Two-Term Spectral Asymptotics for the Dirichlet Pseudo-Relativistic Kinetic Energy Operator on a Bounded Domain
M.Sc. (Mathematics and Theoretical Physics, LMU and TU Munich), 2010 - 2013
Thesis: Semiclassical quantum dynamics via the method of Stationary Phase for a rigorous approach to Feynman Path Integrals
B.Sc. (Physics, LMU Munich), 2006 - 2010
Thesis: Functional Integral for Gauge Potentials and Bosonic Strings

Honors and Fellowships

Fellow of the Richard-Winter scholarship, 2010 – 2013
Selected to enter the elite graduate program Theoretical and Mathematical Physics at the LMU and TU Munich, 2010
Fellow of the e-fellows.net online scholarship and career network, 2006 – present
Ferry Porsche Prize "for outstanding performance in Mathematics and Physics", 2005
2nd prize at Jugend forscht, 2000 (German youth science competition)

Teaching Experience

Institute of Neural Information Processesing, Ulm University, 2016 - present
Supervising a team of interdisciplinary PhD students, supervising undergraduate students, giving lectures and exercise sessions.
Department of Mathematics, LMU Munich, 2013 - 2016
Teaching mathematics and informatics students in the following courses as a teaching assistant:
- Functional Analysis, Prof. Thomas Østergaard Sørensen PhD, Summer 2016
- Functional Analysis II, Prof. Thomas Østergaard Sørensen PhD, Winter 2015/2016
- Functional Analysis, Prof. Martin Fraas PhD, Summer 2015
- Functional Analysis II, Prof. Thomas Østergaard Sørensen PhD, Winter 2014/15
- Functional Analysis, Prof. Dr. Peter Müller, Summer 2014
- Analysis for Computer Scientists and Statisticians, Prof. Dr. Peter Otte, Winter 2013/2014
- Topology and Differential Calculus in Several Variables, Prof. Dr. Merkl, Summer 2013
Faculties of Physics, Mathematics, Informatics, and Statistics, LMU Munich, 2007 - 2012
Tutor and marker as a working student for the following lectures:
- Mathematical Quantum Mechanics, Prof. László Erdös, PhD, Winter 2012
- Measure and Integration Theory, Dr. Ralf Gerkmann, Summer 2012
- Functional Analysis II, Prof. Thomas Østergaard Sørensen PhD, Winter 2011
- Mathematics II for Natural Scientists, PD Dr. Doris Jakubassa-Amundsen, Summer 2011
- Linear Algebra, Prof. Dr. Peter Pickl, Summer 2011
- Mathematics I for Natural Scientists, PD Dr. Doris Jakubassa-Amundsen, Winter 2010
- Mathematics II for Natural Scientists, PD Dr. Doris Jakubassa-Amundsen, Summer 2010
- Classical Mechanics, Prof. Dr. Jan von Delft, Summer 2009
- Quantum Mechanics, Prof. Dr. Stefan Kehrein, Winter 2007
Faculty of Physics, LMU Munich, 2006-2008
Shared lecture notes partly written as a working student:
- Elektrodynamik, Prof. Dr. Gerhard Buchalla, Summer 2008
- Quantenmechanik, Prof. Dr. Stefan Kehrein, Winter 2007
- Theoretische Mechanik, Prof. Dr. Jan von Delft, Summer 2007
- Rechenmethoden, Bernhard Emmer, Winter 2006