Neurosymbolic integration

We develop neurosymbolic integration both at the conceptual level as well as in several application areas, including chemistry, material science and human behaviour. We use well-developed ontological background knowledge in order to enhance the performance of machine learning methods, e.g. by using a semantic loss function or neuro-fuzzy rules. We have developed a novel neurosymbolic architecture for ontological classification of structured entities and of ontology extension.

Current projects

StrOntEx (Ontology Extension by Automated Learning and Reasoning from Structured Entities) funded by DFG
DigiMatUs Digitalisierung der Materialforschung an Dünnschichtmaterialien am Beispiel von hochauflösenden piezoelektrischen Ultraschallsensoren (Digitisation of materials research on thin-film materials using the example of high-resolution piezoelectric ultrasonic sensors) funded by BMBF

Completed projects

Human Behaviour Change Advancing behavioural science through AI and machine learning (external cooperation)

Publications

Martin Glauer, Till Mossakowski, Fabian Neuhaus, Adel Memariani and Janna Hastings. Neuro-symbolic semantic learning for chemistry. In: P. Hitzler, M. K. Sarker and A. Eberhart, editors, A Compendium of Neuro-Symbolic Artificial Intelligence. Frontiers in Artificial Intelligence and Applications vol. 369, chapter 21, pages 460 - 484. IOS press, 2023.

Martin Glauer, Fabian Neuhaus, Till Mossakowski and Janna Hastings. Ontology Pre-training for Poison Prediction.
CoRR, abs/2301.08577, 2023. German conference on artificial intelligence 2023, to appear. Best paper award.

Martin Glauer, Adel Memariani, Fabian Neuhaus, Till Mossakowski and Janna Hastings. Interpretable Ontology Extension in Chemistry. Semantic Web journal, 2022. Special Issue on The Role of Ontologies and Knowledge in Explainable AI, to appear

Till Mossakowski. Modular design patterns for neural-symbolic integration: refinement and combination.
In: A. d'Avila Garcez and E. Jiménez-Ruiz, editors, 16th International Workshop on Neural-Symbolic Learning and Reasoning (NeSy), volume 3212, series CEUR Workshop proceedings, pages 192-201. 2022.

Martin Glauer, Robert West, Susan Michie, Janna Hastings: ESC-Rules: Explainable, Semantically Constrained Rule Sets. In: A. d'Avila Garcez and E. Jiménez-Ruiz, editors, 16th International Workshop on Neural-Symbolic Learning and Reasoning (NeSy), volume 3212, series CEUR Workshop proceedings, pages : 94-103, 2022.

Janna Hastings, Martin Glauer, Adel Memariani, Fabian Neuhaus and Till Mossakowski. Learning chemistry: exploring the suitability of machine learning for the task of structure-based chemical ontology classification. Journal of Cheminformatics, 13(1):23, 2021.

Adel Memariani, Martin Glauer, Fabian Neuhaus, Till Mossakowski and Janna Hastings. Automated and explainable ontology extension based on deep learning: A case study in the chemical domain.
In: R. Confalonieri, O. Kutz and D. Calvanese, editors, International Workshop on Data meets Applied Ontologies in Explainable AI (DAO-XAI 2021), volume 2998, series CEUR Workshop Proceedings. http://ceur-ws.org/Vol-2998/, 2021.

Till Mossakowski, Razvan Diaconescu and Martin Glauer. Towards Fuzzy Neural Conceptors. IfCoLog Journal of Logics and their Applications, 6(4):725-744, 2019.