Research Portfolio
My research addresses new data science and machine learning methods that help us to understand collective behavior in complex systems with many interacting elements. My approach is quantitative, interdisciplinary, and data-driven, combining methods from machine learning, statistical modelling, graph theory, and the physics of complex systems.
Foundations of Network Science and Graph Learning
My group advances theoretical foundations of network science, graph learning and visualization. Exemplary contributions include:
- A-D Almasan, S Shvydun, I Scholtes, P Van Mieghem: Generating Temporal Contact Graphs Using Random Walkers, IEEE Transactions on Network Science and Engineering, January 2025
- C Blöcker, C Tan, I Scholtes: The Map Equation Goes Neural: Mapping Network Flows with Graph Neural Networks, NeurIPS, December 2024
- C Blöcker, J Smiljanić, I Scholtes, M Rosvall: Similarity-based Link Prediction from Modular Compression of Network Flows, PMLR, December 2022
- LV Petrović, I Scholtes: Learning the Markov order of paths in graphs, The Web Conference, April 2022
- V Perri and I Scholtes: HOTVis: Higher-Order Time-Aware Visualisation of Dynamic Graphs, Graph Drawing, September 2020
- R Pfitzner, I Scholtes, A Garas, CJ Tessone, F Schweitzer: Betweenness Preference: Quantifying Correlations in the Topological Dynamics of Temporal Networks, Physical Review Letters, May 2013
A specific focus of my work is to understand how temporal interaction patterns influence the causal topology of complex systems, i.e. which elements can directly or indirectly influence each other. Over the past years, my group used these insights to develop causality-aware graph learning methods:
- F Heeg, J Sauer, P Mutzel, I Scholtes: Weisfeiler and Leman Follow the Arrow of Time: Expressive Power of Message Passing in Temporal Event Graphs, Preprint, May 2025
- F Heeg, I Scholtes: Using Causality-Aware Graph Neural Networks to Predict Temporal Centralities in Dynamic Graphs, NeurIPS, December 2024
- L Qarkaxhija, V Perri, I Scholtes: De Bruijn goes Neural: Causality-Aware Graph Neural Networks for Time Series Data on Dynamic Graphs, PMLR, December 2022
- R Lambiotte, M Rosvall, I Scholtes: From Networks to Optimal Higher-Order Models of Complex Systems, Nature Physics, March 2019
- I Scholtes: When is a Network a Network? Multi-Order Graphical Model Selection in Pathways and Temporal Networks, SIGKDD, August 2017
- I Scholtes, N Wider, R Pfitzner, A Garas, C Tessone, F Schweitzer: Causality-driven slow-down and speed-up of diffusion in non-Markovian temporal networks, Nature Communications, September 2014
Applications in biological, physical, and social systems
Apart from foundational works on artificial intelligence, my group applies these methods to answer open questions in biology, physics, social science, humanities, and engineering, such as:
- How do interactions between cells in biological tissue influence gene expression? [learn more]
- How do collaboration structures in software teams influence the productivity of developers? [learn more]
- How do temporal collaboration patterns of developers influence the success and failure of Open Source Communities? [learn more]
- How can we monitor and improve the evolution of complex software architectures? [learn more]
- How do collaboration patterns influence the success of scientists? [learn more]
- What insights can deep learning provide on the structure of school curricula? [learn more]
- How can we use computational methods to analyze narrative structures in literature? [learn more]
- How can deep learning speed up computationally expensive simulations in plasma physics?
I have published in top-tier interdisciplinary journals such as Nature Physics, Nature Communications, and Physical Review Letters, as well as in A* computer science venues like ICSE, SIGKDD, The Web Conference, and NeurIPS.
Third-party Funded Projects
My research in the areas outlined above was selected for major funding awards with a volume of more than 3.2 mio EUR.
- co-PI in EU Doctoral Training Network GenAIDE on generative AI in industrial design engineering
total volume EUR 3.717 mio, own share EUR 272,992
2025 - 2029 - co-PI in BMBF Project COMFORT: Compression methods for robustness and transferability
total volume EUR 1.98 mio, own share EUR 294,052
2024 - 2027 - PI in BMBF project Software Campus 3.0
total volume EUR 137,000
2024 - 2026 - co-PI in BMBF project TissueNet: Deep Learning of inter- and intra-cellular network dynamics in CAR-T therapy
total volume EUR 1.148 mio, own share EUR 227,286
2023 - 2026 - PI in industry project on human-AI collaboration, funded by the Honda Research Institute Europe GmbH.
total volume: EUR 432,946
2020 - 2024 - PI in SNF project Next-Generation Network Analytics for Time Series Data
total volume EUR 1.65 mio
2018 - 2024 - co-PI in SERI project The interplay between social information processing and network-based information ranking funded by Swiss State Secretariat for Education, Research and Innovation (SERI)
own share EUR 194,816
2015 - 2018 - co-PI in knowledge transfer project The influence of Interaction Patterns on Success in Socio-Technical Systems funded by the MTEC Foundation at ETH Zurich
own share EUR 43,291
2014 - 2015
Software, Tools, and Technology Transfer
My group actively engages in the transfer of our research into practice, developing open source software and tools that help researchers as well as practitioners from industry.
- I am creator of the Open Source temporal graph learning package pathpyG, which is actively developed in my group. It facilitates GPU-accelerated causality-aware deep learning in time series data on complex networks.
- In the context of our third-party funded project Software Campus 3.0, we work with an industry partner to develop the online analytics platform gitBoard, which provides an AI-based early warning of organizational risks in large software development projects, thus helping to mitigate risks in software supply chains.
- With Dr. Christoph Gote, I created the Open Source data mining package git2net. It allows to use open data from git repositories to infer fine-grained temporal interaction networks between software developers. The package received a Special Mention in the Free and Open Source Software (FOSS) Impact Paper Award.
- From 2004 to 2008 I was developer of the Event Monitoring Service (Emon), a large-scale data distribution system that is currently used to analyze particle collision data from the ATLAS detector at CERN's Large Hadron Collider Experiment.