Smart Seminar: Equations from data: methodologies and limits

Angelo Vulpiani

Presenter: Angelo Vulpiani
Thursday 30th November, 17:30
Location: Maxwell conference room” – 5th floor of the Electronic Department
Politecnico di Torino – Corso Castelfidardo 42, Torino

 

Abstract

In many situations, it is not trivial write evolutionary equations starting from general theories. A typical situation is present while analyzing ecological models.
In case the relevant variables used to describe a particular phenomenon are unknown, and we have only a single historic series of observable, we have to use a methodology created by Takens.
This methodology, under specific hypotheses, allows “reconstructing the space of the phases”. In principle, this methodology allows finding (at least approximately) the evolutionary equations.
However, one of the results of the ergodic theory (Kac’s lemma), shows that exploit this kind of methodologies has several difficulties when the number of variables exceeds a particular threshold (around 5). This limit seems a problem, especially in the Big Data era.

Bio

Angelo Vulpiani is a Professor of Theoretical Physics at the University of Rome Sapienza, and is a Fellow of the Institute of Physics. His research interests are statistical mechanics, dynamical systems, turbulence, transport and reaction-diffusion in fluids.

 

Attachment:
Vupliani SmartSeminar.pdf

 

One thought on “Smart Seminar: Equations from data: methodologies and limits”

Leave a comment

Your email address will not be published. Required fields are marked *