Binary interaction methods for high dimensional global optimization and machine learning


Alessandro Benfenati, Giacomo Borghi, Lorenzo Pareschi (7/5/2021 preprint arXiv:2105.02695)

In this work we introduce a new class of gradient-free global optimization methods based on a binary interaction dynamics governed by a Boltzmann type equation. In each interaction the particles act taking into account both the best microscopic binary position and the best macroscopic collective position. In the mean-field limit we show that the resulting Fokker-Planck partial differential equations generalize the current class of consensus based optimization (CBO) methods. For the latter methods, convergence to the global minimizer can be shown for a large class of functions.

Anisotropic Diffusion in Consensus-based Optimization on the Sphere


Massimo Fornasier, Hui Huang, Lorenzo Pareschi, Philippe Sünnen (1/4/2021 preprint arXiv:2104.00420)

In this paper we are concerned with the global minimization of a possibly non-smooth and non-convex objective function constrained on the unit hypersphere by means of a multi-agent derivative-free method. The proposed algorithm falls into the class of the recently introduced Consensus-Based Optimization. In fact, agents move on the sphere driven by a drift towards an instantaneous consensus point, which is computed as a convex combination of agent locations, weighted by the cost function according to Laplaces principle, and it represents an approximation to a global minimizer. The dynamics is further perturbed by an anisotropic random vector field to favor exploration.

Mean-field control variate methods for kinetic equations with uncertainties and applications to socio-economic sciences

 

Lorenzo Pareschi, Torsten Trimborn, Mattia Zanella (4/2/2021 preprint arXiv:2102.02589

In this paper, we extend a recently introduced multi-fidelity control variate for the uncertainty quantification of the Boltzmann equation to the case of kinetic models arising in the study of multiagent systems. For these phenomena, where the effect of uncertainties is particularly evident, several models have been developed whose equilibrium states are typically unknown. In particular, we aim to develop efficient numerical methods based on solving the kinetic equations in the phase space by Direct Simulation Monte Carlo (DSMC) coupled to a Monte Carlo sampling in the random space.

Modeling and simulating the spatial spread of an epidemic through multiscale kinetic transport equations

 


Walter Boscheri, Giacomo Dimarco, Lorenzo Pareschi (18/12/2020, preprint arXiv:2012.10101, to appear in Math. Mod. Meth. App. Scie.)

In this work we propose a novel space-dependent multiscale model for the spread of infectious diseases in a two-dimensional spatial context on realistic geographical scenarios. The model couples a system of kinetic transport equations describing a population of commuters moving on a large scale (extra-urban) with a system of diffusion equations characterizing the non commuting population acting over a small scale (urban). The modeling approach permits to avoid unrealistic effects of traditional diffusion models in epidemiology, like infinite propagation speed on large scales and mass migration dynamics. A construction based on the transport formalism of kinetic theory allows to give a clear model interpretation to the interactions between infected and susceptible in compartmental space-dependent models.

From particle swarm optimization to consensus based optimization: stochastic modeling and mean-field limit

 

Sara Grassi, Lorenzo Pareschi (10/12/2020 preprint arXiv:2012.05613), Math. Mod. Meth. App. Scie. to appear

In this paper we consider a continuous description based on stochastic differential equations of the popular particle swarm optimization (PSO) process for solving global optimization problems and derive in the large particle limit the corresponding mean-field approximation based on Vlasov-Fokker-Planck-type equations. The disadvantage of memory effects induced by the need to store the local best position is overcome by the introduction of an additional differential equation describing the evolution of the local best.