20 April 2020 - Vaishak Belle

Slides 

https://d.pr/free/f/CskhKi

Speaker

Vaishak Belle

Title

Six perspectives on logic & learning (in infinite domains)

Abstract

The unification of low-level perception and high-level reasoning is a long-standing problem in artificial intelligence, and among other approaches, the integration of logic and learning potentially offers the most general solution to that problem. Although there has been considerable progress on this integration, models in practise continue to make the finite domain assumption, and so models are essentially propositional, programs are loop-free, and so on. In this talk, we discuss a number of different ways in which the infinite is embraced. In recent work, for example, we have looked at the problems of inference and (parameter and structure) learning in continuous domains, that is, where logical atoms model continuous properties. In other work, we report on the synthesis of plans with loops in the presence of probabilistic nondeterminism. Finally, we touch on proposals for declaratively modelling logical reasoning, probabilistic inference and learning problems in continuous domains.

This is a survey talk, drawn from numerous papers, given at the recent dagstuhl seminar on Logic and Learning. Details can be found at: https://drops.dagstuhl.de/opus/volltexte/2020/11842/pdf/dagrep_v009_i009_p001_19361.pdf