Machine learning of structure-property-relationships

• Patterns in chemical sequences that control, for instance, the interaction of copolymers with lipid membranes.
• Hidden physical variables in chemical space revealed by „trans-encoder“ neural networks [1].
• Inverse problems and artificial intelligence in soft materials design.

Static and dynamic conformation patterns

• Unsupervised learning of classes of polymer conformations and collective ordering.
• Precursor patterns for crystal nucleation in entangled polymer melts.

Data-driven coarse-graining of simulation models

• Machine learned implicit solvents and their potential for larger scale simulations of self-assembly.
• Transfer learning of physical patterns between different levels of coarse-graining.

Selected publications:

[1] M. Werner, ACS Macro Letters 10 (2021) 1333-1338.

[2] M. Werner, Y. Guo, V.A. Baulin, npj Computational Materials 6 (2020) 72.