Inference in soft matter: from trajectories to fields
Lead Research Organisation:
University of Cambridge
Department Name: Applied Maths and Theoretical Physics
Abstract
The goal of this project is to apply methods of Bayesian inference to the problem of parameter inference and model selection for experiments in soft matter physics. These may be trajectories of Brownian particles (from which one might want to infer particle properties like radius, mass, or charge) or time-dependent field configurations recorded through a noisy sensor (from which one might want to reconstruct the original field). A related goal is to understand how one may distinguish between reversible and irreversible dynamics through information contained in trajectories or field configurations alone. Applications would include active particle systems, particle imaging velocimetry (PIV), and the dynamics of probes in a complex fluid or soft solid.
Organisations
People |
ORCID iD |
| Günther Turk (Student) |
Studentship Projects
| Project Reference | Relationship | Related To | Start | End | Student Name |
|---|---|---|---|---|---|
| EP/N509620/1 | 01/10/2016 | 30/09/2022 | |||
| 2089780 | Studentship | EP/N509620/1 | 01/10/2018 | 31/12/2022 | Günther Turk |
| EP/R513180/1 | 01/10/2018 | 30/09/2023 | |||
| 2089780 | Studentship | EP/R513180/1 | 01/10/2018 | 31/12/2022 | Günther Turk |
| Description | Microhydrodynamics: We have found inherently positive-definite analytic expressions for the mobility and friction tensors, determining the motion of a mesoscopic Brownian particle of spherical shape near a plane interface between two fluids to high accuracy. Inference in soft matter: We have generalised and extended the pre-existing theoretical framework for Bayesian inference in soft matter systems, in particular, for systems described or approximated by the Ornstein-Uhlenbeck process. Together with our collaborators we are planning to apply this to experimental data. Epidemiological modelling: For much of 2020 most of the Soft Matter group at DAMTP joined the Rapid Assistance in Modelling the Pandemic (RAMP) initiative, tackling the Covid-19 pandemic. We developed, and are still maintaining and expanding, PyRoss, a state-of-the-art Bayesian machine learning platform for epidemiology. |
| Exploitation Route | Microhydrodynamics: These results, once published, should be of interest in colloidal physics to both, experiment and simulation of eg Brownian dynamics. Inference in soft matter: Much remains to be done, which includes testing our theory on experimental data, and further developing not just the theoretical framework, but also the efficient numerics. We expect the results of this course of work to be of future use to other researchers in our field via an open source inference software, especially suited for systems in soft matter. PyRoss: This open source state-of-the-art Bayesian machine learning platform for epidemiology is being actively used by members of the developing group and modellers from elsewhere in order to infer crucial information from national data on Covid-19 (not restricted to the UK) and potentially be able to make real-time policy-influencing predictions. |
| Sectors | Digital/Communication/Information Technologies (including Software),Healthcare,Manufacturing, including Industrial Biotechology,Other |
| URL | https://github.com/rajeshrinet/pyross |
| Description | PyRoss has been used to infer important information on the Covid-19 pandemic from national data of England and Wales, Germany and France. PyRoss is a highly flexible software package that offers an integrated platform for inference, forecasts and non-pharmaceutical interventions in structured epidemiological compartment models. We hope it will be used by many people around the globe interested in epidemiological modelling, in order to help better understand the current pandemic that has been affecting lives worldwide. |
| First Year Of Impact | 2020 |
| Sector | Communities and Social Services/Policy |