Postdoc position in Physics-aware symbolic regression for dynamic systems modeling

Symbolic regression (SR) is a method that automatically generates models as analytic free-form formulas from data. SR has been successfully used in many nonlinear modeling tasks with quite impressive results. SR has several advantages over other data-driven modeling methods. Contrary to (deep) neural networks, which belong to data-hungry approaches, SR can construct good models even from very small training data sets. It is also suitable for incorporating prior knowledge about the desired properties of the modeled system, thus allowing the construction of both precise and physically plausible models.

Though the SR has been studied for around three decades, there are still many open challenges in the field. One of them is solving the problems that lead to implicit equations, meaning that the target variable is not directly present in the training data as a separate value. A particular case is a problem where the model sought represents possibly multiple partial differential equations (PDEs) governing the unknown dynamics of the observed system. This project aims to contribute to this area. The research should focus on developing new SR methods based on genetic programming, neural networks or transformers. In any case, a mechanism to incorporate prior knowledge should be considered. Further, special attention will be paid to investigating the possibilities of using large language models to assist with defining the underlying physics constraints for the studied system.

We seek a motivated Postdoc candidate with a PhD degree in computer science, engineering, or a related field with a strong interest in robotics, artificial intelligence, or human-robot interaction. We expect the candidates to perform excellent research, become part of the world's research communities in your field, and publish in first-tier scientific conferences and journals. This requires critical thinking, creativity, and excellent communication skills in English. The work requires developing code in a Unix environment with Python and/or C++. Experience with ROS, automated planning, Python AI frameworks, and scientific computing is a plus.

We offer the opportunity to do scientifically challenging research on a fully funded position (1-year contract with the possibility of an extension of up to 4 years), including full social and health insurance. Work with an experienced international team and lab with several HRC workspaces within the CIIRC CTU in Prague. Live in one of the top cities to live in (cf. Time Out Magazine index for 2022). The Dejvice Campus features a range of student amenities, such as the National Library of Technology, and a laid-back atmosphere with cafes and other social hangout places. Other benefits include a competitive salary, 30 days paid leave, and a family-friendly environment (Children’s corner, kindergarten, and elementary school operated by the Czech Technical University in Prague).

Cooperation with partners within the consortium of our existing European project CoreSense (Delft University of Technology, Universidad Politécnica de Madrid, etc.) is planned and encouraged.