A scientific competition led to improved methods for analysing the diffusion of particles
By organising a scientific competition about characterising anomalous diffusion, researchers have learnt how the tools used to analyse the diffusion of microscopic particles can be improved. This is shown in a new study co-written by researchers at the Department of Physics.
Methods for detecting anomalous diffusion, meaning the deviation from Brownian motion that can be found in for example biomolecules inside cells and active matter systems, have been developed using classical statistics. Now, the use of machine learning offers the possibility of more refined tools.
With the aim to assess the available methods to characterise anomalous diffusion, and spur the invention of new approaches, an open scientific competition was launched, The Anomalous Diffusion (AnDi) Challenge.
“The original idea is from Carlo Manzo at the University of Vic. I joined during the initial planning stages of the competition after a student told me about it, and I found the idea extremely interesting,” says Giovanni Volpe, professor at the Department of Physics.
Giovanni Volpe co-organised the competition, with researchers from University of Vic – Central University of Catalunya, Institute of Photonic Sciences in Barcelona, University of Potsdam, and Valencia Polytechnic University.
Aimed to provide an objective assessment of methods
The challenge was held during March–November 2020 and consisted of three main tasks concerning anomalous exponent inference, model classification, and trajectory segmentation. The goal was to provide an objective assessment of the performance of methods to characterise anomalous diffusion from single trajectories.
“For the competition, we let developers build and use their own tools to provide predictions for the common dataset. While this choice limited the methods assessed to those provided by the community, it ensured that those algorithms were properly applied,” says Giovanni Volpe.
Aykut Argun, postdoctor at the Department of Physics and co-writer of the study, formed a team that won five of the in total nine tasks in the challenge. In total, around 20 teams participated.
Using machine learning improves the estimation
The results from the competition lead to a study that is now published, that analyses the results of the community effort and determines that machine learning greatly improves the estimation of the properties of diffusing particles.
“By organising this competition, we have learnt that the tools currently employed to analyse the diffusion of microscopic particles can be improved, especially by using advanced machine learning tools. We will further explore these issues in the second edition of the challenge that we are now organising,” says Giovanni Volpe.
About the research
Title: Objective comparison of methods to decode anomalous diffusion
Scientific journal: Nature Communications