sucksxx
mobile porn
sextube
russiporn
Schlampe Porno
mom sex videos
hindi porn
desi porn
Sharing cum with her BFF asian porn xxx videos bf porn kalebet betovisn betvoy vslot exxenbet milosbet
news

CLabB Seminar, Reconstructing dynamical networks via feature ranking (MG Leguina)

news image

CLabB projects 2019/10/14

Title: Reconstructing dynamical networks via feature ranking

Speaker: Marc Grau Leguina, Wyss Center Postdoctoral Fellow, Department of Neurology, Inselspital, University Hospital Bern, University of Bern, Switzerland

When: October 17th, 2019, 15:00 Where: Room 3.20 3rd floor, Departament de Física de la Matèria Condensada, Facultat de Física, Marti i Franques 1  

Abstract: Empirical data on real complex systems are becoming increasingly available. Parallel to this is the need for new methods of reconstructing (inferring) the structure of networks from time- resolved observations of their node-dynamics. The methods based on physical insights often rely on strong assumptions about the properties and dynamics of the scrutinized network. Here, we use the insights from machine learning to design a new method of network reconstruction that essentially makes no such assumptions. Specifically, we interpret the available trajectories (data) as “features” and use two independent feature ranking approaches—Random Forest and RReliefF—to rank the importance of each node for predicting the value of each other node, which yields the reconstructed adjacency matrix. We show that our method is fairly robust to coupling strength, system size, trajectory length, and noise. We also find that the reconstruction quality strongly depends on the dynamical regime.

Link to the paper: M. G. Leguia, Z. Levnajic, L. Todorovski, and B. Zenko, “Reconstructing dynamical networks via feature ranking,” Chaos 29, 093107 (2019) https://doi.org/10.1063/1.5092170