 |
 |
Olga Egorova, (UoS) When Thursday, December 5, 2pm – 3pm Where S5.20 King's College London, Strand Title. Multi-fidelity statistical modelling approach for organic molecular crystal structure prediction Abstract: Structural polymorphism is a phenomenon whereby a molecule can crystallise into different solid forms. The variety of resulting crystal structures and associated lattice energies leads to the differences in physical and chemical properties – and these can crucially affect, for example, the suitability and safety of a pharmaceutical compound. Discovery of new crystal forms and exploration of the energy landscapes is guided by computational methods for crystal structure prediction (CSP), which form a hierarchy in terms of both accuracy and computational costs. In this work we exploit multi-fidelity Bayesian Gaussian processes modelling to combine different levels of theory to obtain predictions for the highest level and lower the costs. Another objective of interest is identifying robust local energy minima structures, and we are incorporating Bayesian optimisation (as in sequential optimal design) for the purposes of searching for the minima and describing the response surface in the area of each minimum. [Joint work with D. Woods, G. Day, R. Hafizi] ______________________________________________________ Seminar organiser: Kalliopi Mylona [log in to unmask] https://mth.kcl.ac.uk/statistics/ The group runs a regular seminar programme on Thursday afternoons, currently 14.00 in room S5.20 on the Strand Campus (Strand, London WC2R 2LS). You may leave the list at any time by sending the command SIGNOFF allstat to [log in to unmask], leaving the subject line blank.