Physics & QBio Hagoromo Hour: Federica Ferretti, MIT, “An efficient Bayesian inference scheme for flocking models with behavioral inertia”

Event time: 
Wednesday, May 3, 2023 - 11:00am to 12:00pm
Location: 
Bass Center for Molecular and Structural Biology, Room 305 See map
266 Whitney Avenue
New Haven, CT 06511
Speaker/Performer: 
Federica Ferretti, MIT
Event description: 

Bird flocks are a paradigmatic example of active matter systems, whose collective motion can be understood as a symmetry-breaking phenomenon emerging from microscopic alignment interactions. The experimental observation of flocks of European starlings has revealed the need to update traditional Vicsek-like models, used to describe such collective behavior, with the introduction of a behavioral inertia. Inferring this term from experimental data is a fundamental step to test the hypothesis that inertia may control the finite size of real flocks.
In this talk I will consider the first agent-based model that incorporated this new ingredient, the Inertial Spin Model [1], and discuss the challenges imposed by its non-Markovian character on the design of an efficient Bayesian inference scheme for such many-body stochastic process. I will illustrate an algorithm that reduces the corresponding maximum likelihood problem to a linear regression (applicable to a broader class of partially observed diffusion processes) and discuss the pitfalls of a naive use of delay vector embeddings in this context [2,3].
[1] Attanasi et al. “Information transfer and behavioural inertia in starling flocks.” Nature Physics 10 (July 2014), https://doi.org/10.1038/nphys3035
[2] Ferretti et al., “Building general Langevin models from discrete datasets.” Phys. Rev. X 10 (July 2020), https://doi.org/10.1103/PhysRevX.10.031018
[3] Ferretti et al., “Renormalization group approach to connect discrete- and continuous-time descriptions of Gaussian processes.” Phys. Rev. E 105 (April 2022) https://doi/10.1103/PhysRevE.105.044133
Hosts: Michael Abbott (michael.abbott@yale.edu), Isabella Graf (isabella.graf@yale.edu), and Mason Rouches (mason.rouches@yale.edu)

Admission: 
Free