Prof. Dr.
Kristian Kersting
Technische Universität Darmstadt
Computer Science Department and Centre for Cognitive Science
Hochschulstraße 1
64289 Darmstadt
Short info
My Artificial Intelligence and Machine Learning lab would like to make computers learn so much about the world, so rapidly and flexibly, as humans. This poses many deep and fascinating scientific problems: How can computers learn with less help from us and data? How can computers reason about and learn with complex data such as graphs and uncertain databases? How can pre-existing knowledge be exploited? How can computers decide autonomously which representation is best for the data at hand? Can learned results be physically plausible or be made understandable by us? How can computers learn together with us in the loop? To this end, we develop novel machine learning (ML) and artificial intelligence (AI) methods, i.e., novel computational methods that contain and combine for example search, logical and probabilistic techniques as well as (deep) (un)supervised and reinforcement learning methods.
Open Science
Sum-Product-Attention Networks: Leveraging Self-Attention in Probabilistic Circuits.
arXiv preprint
Structural Causal Interpretation Theorem.
arXiv preprint
Articles
Multifusion: Fusing pre-trained models for multi-lingual, multi-modal image generation.
Advances in Neural Information Processing Systems, 36.
SEGA: Instructing text-to-image models using semantic guidance.
Advances in Neural Information Processing Systems, 36.
Illume: Rationalizing vision-language models through human interactions.
An object numbering task reveals an underestimation of complexity for typically structured scenes.
Psychonomic Bulletin & Review, 1-10.
Atman: Understanding transformer predictions through memory efficient attention manipulation.
Advances in Neural Information Processing Systems, 36.
A typology for exploring the mitigation of shortcut behaviour.
Nature Machine Intelligence, 5(3), 319-330.
Revision Transformers: Instructing Language Models to Change their Values. In
Proceedings of the 26th European Conference on Artificial Intelligence (ECAI).
Safe latent diffusion: Mitigating inappropriate degeneration in diffusion models. In
Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition(pp. 22522-22531).
Large pre-trained language models contain human-like biases of what is right and wrong to do.
Nature Machine Intelligence, 4(3), 258-268.
α ILP: thinking visual scenes as differentiable logic programs.
Machine Learning, 112(5), 1465-1497.
Scalable Neural-Probabilistic Answer Set Programming.
Journal of Artificial Intelligence Research (JAIR) 78:579–617.
RECOWNs: Probabilistic Circuits for Trustworthy Time Series Forecasting.
Proceedings of the 4th Workshop on Tractable Probabilistic Modeling (TPM 2021).
Modelling dataset bias in machine-learned theories of economic decision-making.
Nature Human Behaviour, 1-13.
Generative Clausal Networks: Relational Decision Trees as Probabilistic Circuits.
Proceedings of the 30th International Conference on Inductive Logic Programming (ILP).
Characteristic Circuits.
Advances in Neural Information Processing Systems, 36.
Whittle networks: A deep likelihood model for time series.
International Conference on Machine Learning (PMLR), 12177-12186.
Leveraging Probabilistic Circuits for Nonparametric Multi-Output Regression.
Proceedings of the 37th Conference on Uncertainty in Artificial Intelligence (UAI 2021).
