Nous sommes heureux de vous annoncer le lancement des séminaires

mensuels du GDR IA, qui verront intervenir chaque mois un orateur sur

un sujet différent, proposés par les groupes de travail du GDR ou les

membres du comité scientifique.

Le but de ces exposés est de mettre en lumière, au sein d’un exposé

accessible à la plus grande partie de la communauté, les domaines

variés couverts par le GDR et ses activités. Si la langue des slides

et du discours sont laissés au choix de l’orateur, notre préconisation

par défaut sont des slides en anglais avec le libre choix de la langue

utilisée.

**Prochain séminaire:**

### 19/01/22 – 11h: Meghyn Bienvenu

#### Titre: **Ontology-Mediated Query Answering: Using Rules and Reasoning to Get More from Data**

Ontology-mediated query answering (OMQA) is a promising approach to data access and integration that has been extensively studied by the knowledge representation & reasoning and database communities. Ontologies are used to enrich and unify the vocabulary of data sources, allowing users to formulate their queries in a more familiar vocabulary which abstracts from the specific way data is stored, and to specify domain knowledge, which can be exploited by reasoning algorithms to infer implicit information and return more complete query results. While the addition of an ontology brings significant benefits, it also renders the query answering task more computationally involved. This has led both to theoretical studies aimed at understanding the complexity of OMQA for a range of ontology and query languages, as well as the development of new algorithmic techniques. The aim of this talk will be to provide a gentle introduction to OMQA, while also highlighting some recent results and research directions.

# Liste des séminaires (passés et à venir, ordre anti-chronologique):

### 6/4/22 – 11h : Leila Amgoud

#### Titre: TBA

### 23/03/22 – 11h : Hervé Isambert

#### Titre: Learning reliable causal networks from multivariate information in biological and clinical data.

### 02/02/22 – 11h: Stéphane Canu

#### Titre: AI in cars

### 19/01/22 – 11h: Meghyn Bienvenu

#### Titre: **Ontology-Mediated Query Answering: Using Rules and Reasoning to Get More from Data**

Ontology-mediated query answering (OMQA) is a promising approach to data access and integration that has been extensively studied by the knowledge representation & reasoning and database communities. Ontologies are used to enrich and unify the vocabulary of data sources, allowing users to formulate their queries in a more familiar vocabulary which abstracts from the specific way data is stored, and to specify domain knowledge, which can be exploited by reasoning algorithms to infer implicit information and return more complete query results. While the addition of an ontology brings significant benefits, it also renders the query answering task more computationally involved. This has led both to theoretical studies aimed at understanding the complexity of OMQA for a range of ontology and query languages, as well as the development of new algorithmic techniques. The aim of this talk will be to provide a gentle introduction to OMQA, while also highlighting some recent results and research directions.

### 15/12/21 – 11h: Meltem Ozturk

#### Titre: Axiomatic and Algorithmic elements of the Social Ranking Problem and its relation with ordinal power indices

In this talk, I will present some recent studies on *social ranking* from axiomatic and computational point of views and show it’s relation with power indices.

In many domains, a number of works have been devoted to ranking individuals/objects based on the performance of the groups formed by them. Besides game theory, arguably the field which has dealt with this question the most, we can think of networks (the influence of someone in a social network), belief merging/revision (the responsibility of a formula in the inconsistency of a belief base), multicriteria decision aiding (the impact and synergy of some criteria), machine learning (selecting best features), argumentation (influence of an argument in a debate), etc.

Power indices (such as Shapley or Banzhaf indices), introduced in cooperative game theory, deal with this problem in a cardinal way. But many real-life situations do not fit this framework, as in particular we may only have ordinal information about groups of agents. Social ranking proposes a more flexible theory of cooperative interaction situations considering only ordinal comparisons between groups.

In this talk, I will present different social ranking rules, show their differences based on their axiomatizations. I will also point out some computational results (such as manipulability of such rules). This ordinal approach being quite recent, I will conclude by some open questions.

- La vidéo de cet exposé:

### 24/11/21 – 11h: Régis Sabbadin

#### Titre: Making new out of old in Nash equilibrium computation: Path-following

among polynomial equations systems

Non-cooperative game theory is particularly useful for modeling interactions between competitive agents. In AI, it is the grounding framework of (non-cooperative) multi-agent planning, for example. AI has also contributed new representation frameworks (graphical games, hypergraphical games, Bayesian Action Graph Games…) and new solution algorithms (Multi-agent RL,…) to game theory. In this introductory talk, I will present game components and basic concepts (non-dominated strategies, Nash equilibrium). Then, I will briefly list Nash equilibrium computation approaches and focus on the links between NE computation and polynomial systems solving. I will recast a few recent and (very) old NE computation algorithms in terms of polynomial systems solving. Finally, I will mention that similar polynomial systems representations also apply to NE computation problems in incomplete information and succinctly-expressed games.

### 13/10/21 – **14h** : Didier Dubois

#### Titre: Unified View of Uncertainty Theories – The limited expressiveness of

single probability measures

The variability of physical phenomena and partial ignorance about them

motivated the development of probability theory in the two last

centuries. However, the mathematical framework of probability theory,

together with the Bayesian credo claiming the inevitability of unique

probability measures for representing agents’ beliefs, have blurred

the distinction between variability and ignorance. Modern theories of

uncertainty, by putting together probabilistic and set-valued

representations of information, provide a better account of the

various facets of uncertainty.

- La vidéo de cet exposé:

### 07/07/21 – **15h** : Dominik Peters

#### Titre: Proportional Participatory Budgeting

We study voting rules for participatory budgeting, where a group of

voters collectively decides which projects should be funded using a

common budget. Many city governments around the world now use

participatory budgeting to allow their residents to influence how the

city spends its budget. Most cities handle such elections using the

most intuitive voting rule: greedily filling a knapsack with projects

that received the highest number of votes. We argue that this is a bad

voting rule, because it can ignore the preferences of substantial

minorities. Instead, we advocate the use of voting rules that are

proportional: every voter should have roughly equal influence on the

final budget, and thus different interests should be represented in

proportion to the number of their supporters. We formalize the notion

of proportionality as an axiom and design a simple and attractive

voting rule that satisfies our formal axiom, and that can be evaluated

in polynomial time. We also prove that a large class of voting rules

based on optimization cannot achieve proportional outcomes: this is

surprising because in many other contexts, researchers have found

specific objective functions that lead to fair outcomes.

Based on joint work with Grzegorz Pierczyński and Piotr Skowron

(https://arxiv.org/abs/2008.13276).

- La vidéo de cet exposé:

### 16/06/21 – 11h : Pierre Talbot

#### Titre: Interprétation abstraite de la programmation par contraintes

Résumé: Durant ce séminaire, nous nous intéressons à la fusion de deux domaines de recherche: la programmation par contraintes (PPC) et l’interprétation abstraite. La PPC est un paradigme déclaratif permettant de trouver les solutions d’un système de contraintes. La PPC s’applique à de nombreux domaines tels que l’ordonnancement, les tournées de véhicules ou encore en bioinformatique et composition musicale. L’interprétation abstraite est une théorie pour l’analyse statique de programmes par approximation de leurs sémantiques. Les idées véhiculées par l’interprétation abstraite se sont démontrées applicables en dehors de l’analyse de programmes, notamment pour la PPC. Au lieu d’approximer la sémantique d’un programme, nous pouvons approximer la sémantique d’une formule logique, représentant l’ensemble des solutions d’un problème de contraintes. Un des avantages d’utiliser une telle formalisation est de réutiliser des outils de l’interprétation abstraite pour la combinaison de solveur par contraintes, capturant différents langages de contraintes. Nous motiverons et présenterons plusieurs combinaisons génériques de solveurs par contraintes, afin de résoudre des problèmes d’ordonnancement. Finalement, nous aborderons un sujet de recherche en cours, explorant l’interprétation abstraite pour la parallélisation de la PPC sur GPU. Le séminaire ne requiert pas de prérequis en PPC ni interprétation abstraite.

- Les slides de cet exposé sont ici: http://hyc.io/papers/talbot-gdr-ia-2021.pdf
- La vidéo de cet exposé:

### 26/05/21 – 11h : Anne Siegel

#### Titre: Modelling unconventional biological systems:

dynamical systems and/or reasoning?

Résumé: In the last years, data sciences have shown their interest

(and somehow their limitation) in life sciences to extract information

from multi-scale, incomplete, heterogeneous but somehow interdependent

data. As an important bottleneck, figuring out how data sciences

techniques can assist life sciences to provide explainable evidences

of the limitations of current knowledge in biology and therefore lead

to news discoveries.

An example of such limitations is the study of biological systems in

molecular and cellular biology. Such systems are highly difficult to

model with a single calibrated numerical model: state of the art

methods require a precise understanding of interactions occurring in

the considered species, an information which is not available for

complex organisms (for instance, eukaryotes) and hardly cultivable

bacteria such as those evidenced in microbiomes.

In this talk, we will introduce a strategy to assist biological

discovery by identifying features of large-scale unconventional

biological systems at the molecular scale despite lacks of data. We

rely on the different reasoning modes (projection, union,

intersection) combinatorial optimization) of Answer Set Programming,

on its semantics of negation and on its combinatorial problem

optimization feature to search for new functionalities of biological

systems. In practice, we use this formalism to reason on

over-approximations of the biological system response with

steady-states of Boolean networks, in order to identify biological

candidates (enzymes, biological functions, or species) involved in

targeted biological features.

We will illustrate this approach on the emerging field of systems

ecology, which aims at understanding interactions between a consortium

of microbes and a host organism, more precisely by discussing the

putative role of host-bacterial interactions in an algal system.

- La vidéo de cet exposé:

### 28/04/21 – 11h : Joao Marques-Silva

#### Titre: Formal Reasoning Methods in Explainable AI

Résumé: The forecasted applications of machine learning (ML) in safety

critical applications hinge on systems that are robust in their

operation and that can be trusted. This talk overviews recent efforts

on applying automated reasoning tools in explaining non-interpretable

(black-box) ML models. Moreover, the talk details the computation of

rigorous explanations of black-box models, and how these serve for

assessing the quality of widely used heuristic explanation approaches.

The talk also covers important properties of rigorous explanations,

namely duality properties between different kinds of explanation.

Finally, the talk briefly overviews ongoing work on mapping tractable

explainability.

- La vidéo de cet exposé:

### 07/04/21 – 11h : Loïc Paulevé

#### Titre: Symbolic learning of ensembles of Boolean networks predictive for cell

fate decision

Résumé: In this talk, I’ll give a global overview of challenges related to the

learning of dynamical models of cellular differentiation processes.

I’ll present on-going work mixing formal methods for modelling dynamical

systems and model synthesis from constraints on expected emerging

behaviors. This will be illustrated in the scope of the synthesis and

analysis of ensembles of Boolean automata networks using Answer-Set

Programming from data relating to the structure of the network and

observations of the system.

- La vidéo de cet exposé: