Papers by Jean-christophe Janodet
Springer eBooks, 2013
Generalized maps describe the subdivision of objects in cells, and incidence and adjacency relati... more Generalized maps describe the subdivision of objects in cells, and incidence and adjacency relations between cells, and they are widely used to model 2D and 3D images. Recently, we have defined submap isomorphism, which involves deciding if a copy of a pattern map may be found in a target map, and we have described a polynomial time algorithm for solving this problem when the pattern map is connected. In this paper, we show that submap isomorphism becomes NP-complete when the pattern map is not connected, by reducing the NP-complete problem Planar-4 3-SAT to it.
La detection d'anomalies est un probleme recurrent en Machine Learning. Des techniques recent... more La detection d'anomalies est un probleme recurrent en Machine Learning. Des techniques recentes cherchenta exploiter le potentiel des GAN (Genera-tive Adversarial Networks) pour detecter les anomalies de facon indirecte, dans l'espace des donnees ; elles se fondent sur l'idee que le generateur ne peut pas recons-truire une anomalie. Nous developpons une approche alternative, basee sur un Encoding Adversarial Network (ANOEAN), qui projette les donnees dans un espace latent, ou la detection d'anomalies se fait de facon directe, en calculant un score. Notre encodeur est appris par adversarial learning, en utilisant deux fonctions de perte, la premiere contraignant l'encodeur a modeliser les donnees normales dans un espace qui suit une distribution gaussienne, et la seconde,a proje-ter des donnees anormales en dehors de cette distribution. Nous conduisons une serie d'experiences sur plu-sieurs bases standard, et montrons que notre approche depasse l'etat de l...
Communications in Computer and Information Science, 2020
Modern and future vehicles are complex cyber-physical systems. The connection to their outside en... more Modern and future vehicles are complex cyber-physical systems. The connection to their outside environment raises many security problems that impact our safety directly. In this work, we propose a Deep CAN intrusion detection system framework. We introduce a multivariate time series representation for asynchronous CAN data which enhances the temporal modelling of deep learning architectures for anomaly detection. We study different deep learning tasks (supervised/unsupervised) and compare several architectures, in order to design an in-vehicle intrusion detection system that fits in-vehicle computational constraints. We conduct experiments with many types of attacks on an in-vehicle CAN using SynCAn Dataset.
La détection d'anomalies est un problème récurrent en Machine Learning. Des techniques récentes c... more La détection d'anomalies est un problème récurrent en Machine Learning. Des techniques récentes cherchentà exploiter le potentiel des GAN (Generative Adversarial Networks) pour détecter les anomalies de façon indirecte, dans l'espace des données ; elles se fondent sur l'idée que le générateur ne peut pas reconstruire une anomalie. Nous développons une approche alternative, basée sur un Encoding Adversarial Network (ANOEAN), qui projette les données dans un espace latent, où la détection d'anomalies se fait de façon directe, en calculant un score. Notre encodeur est appris par adversarial learning, en utilisant deux fonctions de perte, la première contraignant l'encodeur a modéliser les données normales dans un espace qui suit une distribution gaussienne, et la seconde,à projeter des données anormales en dehors de cette distribution. Nous conduisons une série d'expériences sur plusieurs bases standard, et montrons que notre approche dépasse l'état de l'art lorsqu'on utilise 10% d'anomalies lors de l'apprentissage. Mots-clef : Détection d'anomalies, Adversarial Learning, GAN, EAN.
La reconnaissance de formes s'interesse a la detection automatique de motifs dans des donnees... more La reconnaissance de formes s'interesse a la detection automatique de motifs dans des donnees d'entree, afin de pouvoir, par exemple, les classer en categories. La matiere premiere de ces techniques est bien souvent l'image numerique. Cette derniere, dans sa forme la plus courante, est codee sous la forme d'une matrice de pixels. Neanmoins, la question du developpement de representations plus riches se pose. Ainsi, la structuration de l'information contenue dans l'image devrait permettre la mise en evidence des differents objets representes, et des liens les unissant. C'est pourquoi nous proposons de modeliser les images numeriques sous forme de graphes, pour leur richesse et expressivite d'une part, et pour exploiter les resultats de la theorie des graphes en reconnaissance de formes d'autre part. Nous developpons pour cela une methode d'extraction de graphes plans a partir d'images, basee sur le respect de la semantique. Nous montrons qu...
HAL is a multi-disciplinary open access archive for the deposit and dissemination of sci-entific ... more HAL is a multi-disciplinary open access archive for the deposit and dissemination of sci-entific research documents, whether they are pub-lished or not. The documents may come from teaching and research institutions in France or abroad, or from public or private research centers. L’archive ouverte pluridisciplinaire HAL, est destinée au dépôt et a ̀ la diffusion de documents scientifiques de niveau recherche, publiés ou non, émanant des établissements d’enseignement et de recherche français ou étrangers, des laboratoires publics ou privés. On the complexity of Submap Isomorphism and
Lecture Notes in Computer Science, 2004
Powerful methods and algorithms are known to learn regular languages. Aiming at extending them to... more Powerful methods and algorithms are known to learn regular languages. Aiming at extending them to more complex grammars, we choose to change the way we represent these languages. Among the formalisms that allow to define classes of languages, the one of stringrewriting systems (SRS) has outstanding properties. Indeed, SRS are expressive enough to define, in a uniform way, a noteworthy and non trivial class of languages that contains all the regular languages, {a n b n : n ≥ 0}, {w ∈ {a, b} * : |w|a = |w| b }, the parenthesis languages of Dyck, the language of Lukasewitz, and many others. Moreover, SRS constitute an efficient (often linear) parsing device for strings, and are thus promising and challenging candidates in forthcoming applications of Grammatical Inference. In this paper, we pioneer the problem of their learnability. We propose a novel and sound algorithm which allows to identify them in polynomial time. We illustrate the execution of our algorithm throughout a large amount of examples and finally raise some open questions and research directions.
Machine Learning, 2006
Whereas there is a number of methods and algorithms to learn regular languages, moving up the Cho... more Whereas there is a number of methods and algorithms to learn regular languages, moving up the Chomsky hierarchy is proving to be a challenging task. Indeed, several theoretical barriers make the class of context-free languages hard to learn. To tackle these barriers, we choose to change the way we represent these languages. Among the formalisms that allow the definition of classes of languages, the one of string-rewriting systems (SRS) has outstanding properties. We introduce a new type of SRS's, called Delimited SRS (DSRS), that are expressive enough to define, in a uniform way, a noteworthy and non trivial class of languages that contains all the regular languages, {a n b n : n ≥ 0}, {w ∈ {a, b} * : |w| a = |w| b }, the parenthesis languages of Dyck, the language of Lukasiewicz, and many others. Moreover, DSRS's constitute an efficient (often linear) parsing device for strings, and are thus promising candidates in forthcoming applications of grammatical inference. In this paper, we pioneer the problem of their learnability. We propose a novel and sound algorithm (called LARS) which identifies a large subclass of them in polynomial time (but not data). We illustrate the execution of our algorithm through several examples, discuss the position of the class in the Chomsky hierarchy and finally raise some open questions and research directions.
Citeseer
... D. Angluin introduced in (Angluin, 1987) the query learning model, which allows the learner t... more ... D. Angluin introduced in (Angluin, 1987) the query learning model, which allows the learner to ... Al-though membership queries (MQs) and equivalence queries (EQs) have established themselves as the ... This idea was introduced for the first time in (Becerra-Bonache & Yokomori ...
Narrowing constitutes the basis of the operational semantics of modern declarative languages whic... more Narrowing constitutes the basis of the operational semantics of modern declarative languages which integrate functional and logic programming paradigms. Efficient implementations of these languages consider first-order terms as graphs. In this paper, we investigate narrowing in the setting of graph rewriting systems. We take the full advantage of graph structures by allowing maximal sharing of subexpressions and extend the completeness results of the best known narrowing strategies such as needed narrowing or parallel narrowing to the case of constructor-based weakly admissible graph rewrite systems. The resulting graph narrowing strategies share the same optimality results as the corresponding ones for first-order terms and in addition develop shorter derivations.
Modern and future vehicles are complex cyber-physical systems. The connection to their outside en... more Modern and future vehicles are complex cyber-physical systems. The connection to their outside environment raises many security problems that impact our safety directly. In this work, we propose a Deep CAN intrusion detection system framework. We introduce a multivariate time series representation for asynchronous CAN data which enhances the temporal modelling of deep learning architectures for anomaly detection. We study different deep learning tasks (supervised/unsupervised) and compare several architectures, in order to design an in-vehicle intrusion detection system that fits in-vehicle computational constraints. We conduct experiments with many types of attacks on an in-vehicle CAN using SynCAn Dataset.
In this paper, we address the problem of searching for a pattern in a plane graph, i.e., a planar... more In this paper, we address the problem of searching for a pattern in a plane graph, i.e., a planar drawing of a planar graph. To do that, we propose to model plane graphs with 2-dimensional combinatorial maps, which provide nice data structures for modelling the topology of a subdivision of a plane into nodes, edges and faces. We define submap isomorphism, we give a polynomial algorithm for this problem, and we show how this problem may be used to search for a pattern in a plane graph. First experimental results show the validity of this approach to efficiently search for patterns in images.
... admissibles. Jean-Christophe Janodet 1. Les ... cycliques. Nous mettons en évidence uneclasse... more ... admissibles. Jean-Christophe Janodet 1. Les ... cycliques. Nous mettons en évidence uneclasse de graphes cycliques particuliers, les graphes admissibles, pour laquelle nous donnons une preuve de confluence de la réécriture. Concernant ...
Les langages logico-fonctionnels sont des langages de programmation de tres haut niveau permettan... more Les langages logico-fonctionnels sont des langages de programmation de tres haut niveau permettant de definir dans un formalisme unifie des types de donnees, des fonctions et des predicats (relations). Plusieurs propositions de langages logico-fonctionnels ont ete faites mais toutes se restreignent a des calculs bases sur les termes du premier ordre. Cette restriction permet de programmer avec des types abstraits algebriques mais elle rend difficile la manipulation des structures de donnees du monde reel, modelisees sous la forme de graphes cycliques. L'objectif de cette these est donc d'introduire les graphes cycliques comme structure de donnees de base des langages logico-fonctionnels. Pour cela, nous voyons les programmes comme des systemes de reecriture de graphes cycliques et nous etudions les relations de reecriture et de surreduction qu'ils induisent (semantique operationnelle). Une propriete importante de la reecriture concerne la confluence : elle exprime le det...
While some heuristics exist for the learning of graph grammars, few has been done on the theoreti... more While some heuristics exist for the learning of graph grammars, few has been done on the theoretical side. Due to complexity issues, the class of graphs has to be restricted: this paper deals with the subclass of plane graphs, which correspond to drawings of planar graphs. This allows us to introduce a new kind of graph grammars, using a face-replacement mechanism. To learn them, we extend recent successful techniques developed for string grammars, and based on a property on target languages: the substitutability property. We show how this property can be extended to plane graph languages and nally state the rst identication in the limit result for a class of graph grammars, as far as we know.
Büchi automata are used to recognize languages of infinite words. Such languages have been introd... more Büchi automata are used to recognize languages of infinite words. Such languages have been introduced to describe the behavior of real time systems or infinite games. The question of inferring them from infinite examples has already been studied, but it may seem more reasonable to believe that the data from which we want to learn is a set of finite words, namely the prefixes of accepted or rejected infinite words. We describe the problems of identification in the limit and polynomial identification in the limit from given data associated to different interpretations of these prefixes: a positive prefix is universal (respectively existential) when all the infinite words of which it is a prefix are in the language (respectively when at least one is) ; the same applies to the negative prefixes. We prove that the classes of regular ω-languages (those recognized by Büchi automata) and of deterministic ω-languages (those recognized by deterministic Büchi automata) are not identifiable in ...
There are a number of established paradigms to study the learnability of classes of functions or ... more There are a number of established paradigms to study the learnability of classes of functions or languages: Query learning, Identification in the limit, Probably Approximately Correct learning. Comparison between these paradigms is hard. Moreover, when to the question of converging one adds computational constraints, the picture becomes even less clear. We concentrate here on just one class of languages, that of topological balls of strings (for the edit distance), and visit the different learning paradigms in this context. Between the results, we show that surprisingly it is technically easier to learn from text than from an informant.
Quand on cherche a situer l'Inference Grammaticale dans le paysage de la Recherche, on la pla... more Quand on cherche a situer l'Inference Grammaticale dans le paysage de la Recherche, on la place volontiers au sein de l'Apprentissage Automatique, qu'on place lui-meme volontiers dans le champ de l'Intelligence Artificielle. Ainsi, dans leur livre de reference, Laurent Miclet et Antoine Cornuejols preferent-ils parler d'Apprentissage Artificiel plutot que d'Apprentissage Automatique, et consacrent-ils un chapitre complet a l'Inference Grammaticale. C'est l'histoire du Machine Learning qui explique cette hierarchie. Pourtant, en 2010, elle n'est pas toujours facile a justifier : combien de chercheurs dans le domaine du Machine Learning connaissent-ils le paradigme d'identification a la limite ? Et combien de chercheurs en Inference Grammaticale maitrisent-ils la theorie de la regularisation utilisee en optimisation ? Il suffit de suivre des conferences comme ICGI ou ECML pour constater que les communautes sont differentes, tant sur le plan ...
Uploads
Papers by Jean-christophe Janodet