Digits Architectures Logiciels Informatique


Riccietti Elisa : Second order optimization methods for the solution of large scale nonlinear noisy problems

Lundi 16 Mars 2020

**A 9h45** In this talk I will present new second order optimization methods to solve large scale nonlinear problems affected by noise. We distinguish two particular classes of noisy problems depending on the source of the noise. The first class is that of ill-posed least-squares problems with noise on the data. We propose new methods combining trust-region schemes and regularization in order to handle such problems. The second class is that of large scale problems with an expensive objective function, such as those arising in machine learning. We propose to exploit approximations of the objective function of dynamic accuracy to reduce the computational cost of the solution. We design two types of new second order methods (subsampled and multilevel) that are able to deal with the noise introduced by these approximations.