UBC Probability Seminar: Giuseppe Genovese

Hopfield proposed in 1982 as a simple model capable to store and successively retrieve a number of high dimensional patterns, which represents nowadays a cornerstone of the early AI studies. The Hopfield model can be studied from the angle of statistical mechanics and spin glasses. It is remarkably difficult compared to the celebrated Sherrington-Kirkpatrick model, mainly for the presence of pattern retrieval. I will review some old mathematical results about pattern retrieval in Hopfield and related models and present some recent conjectures and developments.