Speaker
Stefano Forte
(Universita' di Milano)
Description
I critically examine issues that arise when applying modern machine learning techniques in the context of precision collider physics, by specifically considering the case of PDF determinations. Questions that I address include: Can we trust the machine learning model to correctly generalize from known examples? How can we validate results, specifically for uncertainties? Can we detect over- and under-learning? Can a machine learning model detect inconsistencies in the data that it is being shown?
