1ere réunion du GT Apprentissage et Raisonnement du GDR IA – IRIT – 15 et 16 Novembre

 

1ere réunion du GT Apprentissage et Raisonnement du GDR IA
IRIT – Salle 03 – 15 et 16 Novembre

 

___

Programme

___

 

Thursday November 15

9h30 -10h
Aim of the working group (general discussion)

10h-12h
Fred Koriche

“Learning to Reason: from theory to practice”

Summary: In learning to reason, the overall goal is to devise prediction algorithms capable of making complex inference tasks, usually performed by
problem solvers. We shall start with the well-studied L2R framework due to Khardon and Roth, and examine both theoretical questions and practical
problems. Notably, we shall explore L2R models and algorithms inspired from convex optimization, kernels methods and deep learning architectures.
For practical applications, we shall discuss about learning models for solving SAT problems, and recent deep models for visual question answering tasks.

14h-16h Steven Schockaert

“From knowledge graph embedding to ontology embedding? An analysis of the compatibility between vector space representations and rules”

16h-16h30 Break

16h30-18h30
Antoine Cornuéjols

“Causality vs. correlation”

Friday November 16

9h-11h
Mathieu Serrurier

“A survey of recent works on explainable AI”

11h30 -12h30
Didier Dubois, Romain Guillaume

“Maximum likelihood under incomplete information: toward a comparison of criteria”

Maximum likelihood is a standard approach to computing a probability distribution that best fits a given dataset. However, when datasets are incomplete or contain imprecise data, a major issue is to properly define the likelihood function to be maximized. This paper highlights the fact that there are several possible likelihood functions to be considered, depending on the purpose to be addressed, namely whether the behavior of the imperfect measurement process causing incompleteness should be included or not in the model, and what are the assumptions we can make or the knowledge we have about this measurement process. Various possible approaches, that differ by the choice of the likelihood function and/or the attitude of the analyst in front of imprecise information are comparatively discussed on examples, and some light is shed on the nature of the corresponding solutions.

14h-15h
Henri Prade

“Revisiting some old works on explanations in reasoning under uncertainty”