AI-based Learning for Physical Simulation
July 19th
Computational physical modeling is a key resource to complement theoretical and experimental methods in modern scientific research and engineering. While access to large amount of data has favored the use of Artificial Intelligence and Machine Learning (ML) techniques to enhance physical simulations, limitations of purely data-driven methods have emerged as concerns their generalization capability and their intelligibility. To overcome these limitations, I propose a hybrid approach that originally combines ML methods and equation-based modeling to significantly improve generalization in small-data scenarios, while guaranteeing the intelligibility of the physical models through the use of symbolic representations. To efficiently handle the computational cost associated with the proposed methods, I will implement them in a new software platform that seamlesslyintegrates automated model learning and high-performance simulation. Thanks to their general-purpose nature, the methods and algorithms developed in this project may be employed in all scientific disciplines and in engineering workflows. In particular, I plan to use them to advance biology and soft robotics by solving challenging modeling tasks.




4/02/2019 14:00-16:00 SALA RIUNIONI (DISG) - 7/02/2019 15:30-17:30 Room 13 - Via Eudossiana,18
A mathematical billiard is the mechanical system of a point particle traveling in a planar domain subject only to elastic collisions with the boundary of the domain (assumed to be infinitely massive). Systems of this kind have wide application in the physical sciences. I will start by giving a light but mathematically rigorous description of a billiard system, trying to show what features of this description lead to the observed properties of the dynamics (e.g., regular, chaotic, etc.). Ergodic theory is the branch of mathematics that studies the properties of a “dynamical system” from a stochastic/probabilistic point of view. It is the main and perhaps only mathematical tool to study systems, such as chaotic billiards, that admit no exact prediction. I will introduce the most basic notions of ergodic theory, together with some elementary examples of dynamical systems, with a view towards the paradigm of the chaotic billiard.


Spontaneous buckling of contractile poroelastic actomyosin sheets
21/11/2018, 15:00-16:00
Shape transitions in developing organisms can be driven by active stresses, notably, active contractility generated by myosin motors. The mechanisms generating tissue folding are typically studied in epithelia. There, the interaction between cells is also coupled to an elastic substrate, presenting a major difficulty for studying contraction induced folding. Here we study the contraction and buckling of active, initially homogeneous, thin elastic actomyosin networks isolated from bounding surfaces. The network behaves as a poroelastic material, where a flow of fluid is generated during contraction. Contraction starts at the system boundaries, proceeds into the bulk, and eventually leads to spontaneous buckling of the sheet at the periphery. The buckling instability resulted from system self-organization and from the spontaneous emergence of density gradients driven by the active contractility. The buckling wavelength increases linearly with sheet thickness. Our system offers a well-controlled way to study mechanically induced, spontaneous shape transitions in active matter.

© Università degli Studi di Roma "La Sapienza" - Piazzale Aldo Moro 5, 00185 Roma