(Machine) learning how to coarse-grain
The workshop will take place as a pure virtual event!
Registered participants will be given access to a Zoom link in due time.
Coarse-grained (CG) models aim at a reduced description of a molecular system, offering not only practical benefits, such as significant computational advantages, but also the means to effectively test what subset of degrees of freedom and interactions are sufficient to describe physical processes of interest [1, 2]. While the last few decades have yielded significant advances in the development of coarse-grained models--from foundational considerations to practical force-field parametrization algorithms and methods--a number of strong assumptions the community makes has plagued its further development. For instance, the persistent description of nonbonded interactions in terms of pairwise functions alone puts a severe bound on the quality of these models, ultimately sacrificing accuracy and transferability.
Machine learning (ML) models--a class of statistical models that systematically improve with increased training data--have recently percolated in many areas of science as a novel powerful tool . While significant developments have been made in the context of applying ML in chemistry and materials science as a way to speed-up computationally-expensive quantum-chemistry calculations [4, 5, 6, 7], the progress for CG models has been much more limited, due in part to a lack of improved computational scaling. While the development of CG force fields, using either kernels  or neural networks [8, 9, 10], have been demonstrated, there is still a need to address more complex systems and computational efficiency.
The purpose of this workshop will be to discuss the current state of the art, some of the challenges that the community is facing in furthering the penetration of ML models in CG simulations, and future perspectives. A number of timely topics will be addressed during the workshop:
Computational efficiency and scaling
Where do we get ahead from traditional force fields? What is the most efficient way of including fundamental symmetries which reduce the amount of data needed ? Can we use data-driven methods to build transferable models across chemistry [11, 18]?
The systematic optimization of mappings from AA to CG
Can ML models link scales more systematically by, i.e., bridging ML ideas such as auto-encoders and dimensionality reduction with coarse-grained variables [12, 13, 14, 15]?
Configurational generators for dense systems
Using ML CG models, can we think of novel ways to generate equilibrium molecular configurations to sample conformational space [16, 17]?
This workshop will be established as a satellite meeting of the TRR146 Conference 2020 “Multiscale simulations of soft matter: New method developments and mathematical foundations” taking place from Sept 30 to Oct 2, 2020. As such, it will be supported both by the Mainz CECAM node, as well as the TRR146 consortia.
Denis Andrienko (Max Planck Institute for Polymer Research) - Organiser
Tristan Bereau (University of Amsterdam) - Organiser