Foto del docente

Daniele Tantari

Professore associato

Dipartimento di Matematica

Settore scientifico disciplinare: MAT/07 FISICA MATEMATICA

Temi di ricerca

Parole chiave: Meccanica Statistica Vetri di spin Reti Neurali Grafi random Finanza Quantitativa Inferenza Statistica Machine Learning

Random financial networks, Dynamic networks, Community detection:

I'm interested in the application of random networks to the study of financial networks: interbank networks, networks of payments, networks of investors, networks of correlations built from time series. In a static perspective I'm interested in the problem of extracting structural informations and inferring macroscopic network structures: communities, hierarchies, core-periphery and bipartite structures. In a dynamical perspective I'm interested both in the study of dynamical processes living on networks (risk/default propagation) than in the modeling of dynamic networks, where links and topology change in time, aimed at finding non-stationary dynamical patterns and at link forecasting.

Neural networks, unsupervised learning:

I'm interested in the study of techniques of unsupervised learning, from Restricted Boltzmann Machines to Deep Learning, from a theoretical point of view. It is possible to study the performances of this techniques and their statistical relevance in relation to the properties of the data training-set and the shape of the learning architecture.

Statistical Physics, Complex Systems, Phase transitions, Emergent behaviors, multipartite systems:

I'm interested in the study of the emergent properties in systems of interacting agents, using techniques from statistical physics of complex systems and critical phenomena. In particular I extended the theory of disordered systems for mean field spin glasses to disordered multipartite systems leveraging on their equivalence with neural networks for associative memory. Multipartite systems are useful for modeling real systems of agents, from biological to social systems.

Ultimi avvisi

Al momento non sono presenti avvisi.