Algo Carè előadása "Interval Predictor Models with Universal Reliability" címmel a SZTAKI-ban

Szeretettel meghívunk minden érdeklődőt a SZTAKI Mérnöki és Üzleti Intelligencia Kutatólaboratóriumának szemináriumára, ahol előadást tart Algo Carè, az olasz University of Brescia mesterséges intelligenciával, gépi tanulással foglalkozó kutatója.

Előadásának címe: „Interval Predictor Models with Universal Reliability”. Az előadás nyelve angol.

Helyszín: SZTAKI, 1111 Budapest, Kende utca 13-17, nagytanácsterem (alagsor)

Időpont: 2019. november 4. hétfő, 15 óra

 Az előadás témája részletesen

In this talk, the problem of estimating a difficult-to-access variable (output) from an accessible one (input) is considered. To this purpose, interval predictor models (IPMs) are introduced to predict the output given the input, based on a limited sample of (independent and identically distributed) data. IPMs will be discussed from a statistical learning perspective, as a fruit of the “algorithmic modelling culture” as opposed to the “data modelling culture”.

In particular, we will focus on a class of IPMs, called Chebyshev layers, that exhibit a special property that we name “universal reliability”: we call “reliability” the probability that a new data point falls inside the predicted interval, and we say that an IPM has “universal reliability” if the distribution of its reliability is always the same irrespective of the data generating distribution.

Some implications, both theoretical and practical, of the universal reliability property will be discussed: namely, implications on the understanding of the accuracy-reliability trade-off, on the role of a-priori knowledge, and on the computation of exact confidence intervals will be considered.

Az előadás alapjául szolgáló tudományos publikáció: S. Garatti, M. C. Campi, A. Carè: On a Class of Interval Predictor Models with Universal Reliability, Automatica, Elsevier, Volume 110, 2019 (https://doi.org/10.1016/j.automatica.2019.108542)

Önéletrajz

Algo Carè received the Ph.D. degree in informatics and automation engineering from the University of Brescia, Italy, in 2013. He is currently a Research Fellow with the Department of Information Engineering at the same university. He spent two years at The University of Melbourne, Melbourne, VIC, Australia, as a Research Fellow in system identification with the Department of Electrical and Electronic Engineering.

Dr. Carè was a recipient of a two-year ERCIM Fellowship in 2016 that he spent at the Institute for Computer Science and Control (SZTAKI), Hungarian Academy of Sciences (MTA), Budapest, Hungary, and at the Multiscale Dynamics Group, National Research Institute for Mathematics and Computer Science (CWI), Amsterdam, The Netherlands.

He received the triennial Stochastic Programming Student Paper Prize by the Stochastic Programming Society for the period 2013 – 2016. His current research interests include data-driven decision methods, system identification, and learning theory.

Az esemény támogatója az EPIC InnoLabs.