BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//University of Liverpool Computer Science Seminar System//v2//EN
BEGIN:VEVENT
DTSTAMP:20260922T101619Z
UID:Seminar-MIF-1403@lxserverM.csc.liv.ac.uk
ORGANIZER:CN=Othon Michail:MAILTO:Othon.Michail@liverpool.ac.uk
DTSTART:20230220T140000
DTEND:20230220T150000
SUMMARY:MIF Series
DESCRIPTION:Nadav Dym: Efficient Invariant Embeddings for 3D point sets.\n\n In many machine learning tasks, the goal is to learn an unknown function which has some known group symmetries. Equivariant machine learning algorithms exploit this by devising architectures (=function spaces) which have these symmetries by construction. Especially relevant examples for chemistry based applications are neural networks for graphs or sets which respect their permutation symmetries, or neural networks for 3D point sets which additionally respect Euclidean symmetries.\nA common theoretical requirement of symmetry based architecture is that they will be able to separate any two objects which are not related by a group symmetry (this property can be used to prove stronger universality results which we will shortly describe). We will review results showing that under very general assumptions such a symmetry preserving separating mapping f exists, and the embedding dimension m can be taken to be roughly twice the dimension of the data. We will then propose a general methodology for efficient computation of such f using random invariants. This methodology is a generalization of the algebraic geometry argument used for the well known proof of phase retrieval injectivity. We will show several applications of this result, and in particular explain how this can be combined with the results in Kurlin's paper (arxiv:2207.08502) to achieve (relatively) efficient separating mappings for 3D point clouds.\nBased on work with Steven J. Gortler, Snir Hordan and Tal Amir and on the papers "Low Dimensional Invariant Embeddings for Universal Geometric Learning?" (arxiv:2205.02956) by Nadav Dym and Steven J. Gortler, and "Complete Neural Networks for Euclidean Graphs" (arxiv:2301.13821) by Snir Hordan, Tal Amir, Steven J. Gortler and Nadav Dym.\n\nhttps://www.csc.liv.ac.uk/research/seminars/abstract.php?id=1403
LOCATION:
END:VEVENT
END:VCALENDAR