Approximation of collective variables by neural networks – Anncolvar
Vojtech Spiwok from UCT
Many methods have been developed to accelerate molecular dynamics simulations. Some of these methods are based on artificial forces applied on certain descriptors of the molecular systems, also known as collective variables. Some potentially useful collective variables are impossible or difficult to calculate on-the-fly during a simulation. To solve this problem we developed a package for approximation of a collective variable using a neural network. This package was demonstrated on metadynamics simulation of mini-protein folding.
Meet us online every friday morning to explore the portfolio of ELIXIR CZ tools and services at https://global.gotomeeting.com/join/508817845 .
Registration is not needed
24. 9. at 9:30 – SBC: an R package for simulation-based calibration and validation of statistical models and algorithms.
1. 10. at 9:30 – to be defined later.
8. 10. at 9:30 – MOLE: Analysis of Protein Channels.
15. 10. at 9:30 – ACC II: Calculation of partial atomic chages.
Please note that online session will be recorded and may be used on ELIXIR CZ website, social media, in ELIXIR CZ e-publications and any other ELIXIR CZ online materials.