Talk by V. Pestov "The curse of dimensionality in some problems of data science"

[Ugo Bruzzo]

============================
SISSA MATHEMATICS SEMINARS
============================

Professor Vladimir Pestov (University of Ottawa, Canada and UFSC, Florianópolis, Brazil)
will give a talk on

The curse of dimensionality in some problems of data science

on Thursday,  July 18th, at 2:30 pm in room 136 (SISSA Building A).

Abstract: The curse of dimensionality in computer science refers to the situation where known algorithms slow down dramatically and become unusable if the dimension of data is too high. We will discuss some geometric and algorithmic aspects of the curse of dimensionality, including open conjectures as well as recent recipes, in the context of such problems as similarity search and statistical machine learning. The mathematical tools used will include asymptotic geometric analysis, descriptive set theory, and mass transformation distance. The talk will be a mixture of theory (presented at an accessible level for a general audience) and some applications, both using toy test datasets from the UCI repository and a real classification problem from genomic data. Some of the work presented is joint with other members of the Data Science group at the University of Ottawa: Hubert Duan, Gaël Giordano, Stan Hatko.

U. Bruzzo