2025Activity report​​​‌Project-TeamVALDA

6.1.1 ProvSQL

• Keywords:​​​‌
  Databases, Provenance, Probability
• Scientific‌ Description:
  ProvSQL is a‌​‌ general and easy-to-deploy provenance​​ tracking and probabilistic database​​​‌ system implemented as a‌ PostgreSQL extension. ProvSQL’s data‌​‌ and query models closely​​ reflect that of a​​​‌ large core of SQL,‌ including multiset semantics, the‌​‌ full relational algebra, and​​ aggregation. A key part​​​‌ of its implementation relies‌ on generic provenance circuits‌​‌ stored in memory-mapped files.​​
• Functional Description:
  The goal​​​‌ of the ProvSQL project‌ is to add support‌​‌ for (m-)semiring provenance and​​ uncertainty management to PostgreSQL​​​‌ databases, in the form‌ of a PostgreSQL extension/module/plugin.‌​‌
• News of the Year:​​
  Compatibility with PostgreSQL 18.​​​‌ Support for PROV-XML output.‌ Partial support of HAVING‌​‌ queries. Support for compiled​​ semirings, including the counting,​​​‌ Boolean, and Why semirings.‌ Basic documentation infrastructure. Temporal‌​‌ semiring and temporal database​​​‌ support. Various minor enhancements​ and bug fixes.
• URL:​‌
  https://­github.­com/­PierreSenellart/­provsql
• Publications:
  hal-05037471,​​ hal-05072212, hal-04930705,​​​‌ hal-04911715, hal-04561331,​ hal-04393781, hal-01672566,​‌ hal-01851538
• Contact:
  Pierre Senellart​​
• Participants:
  Aryak Sen, Pierre​​​‌ Senellart
• Partners:
  Université Grenoble​ Alpes, CNRS, National University​‌ of Singapore

6.1.2 VUS​​

• Name:
  Volume Under the​​​‌ Surface
• Keywords:
  Time Series,​ Anomaly detection, Measures, Performance​‌ measure, Python
• Scientific Description:​​
  Anomaly detection (AD) is​​​‌ a fundamental task for​ time-series analytics with important​‌ implications for the downstream​​ performance of many applications.​​​‌ In contrast to other​ domains where AD mainly​‌ focuses on point-based anomalies​​ (i.e., outliers in standalone​​​‌ observations), AD for time​ series is also concerned​‌ with range-based anomalies (i.e.,​​ outliers spanning multiple observations).​​​‌ Nevertheless, it is common​ to use traditional point-based​‌ information retrieval measures, such​​ as Precision, Recall, and​​​‌ F-score, to assess the​ quality of methods by​‌ thresholding the anomaly score​​ to mark each point​​​‌ as an anomaly or​ not. However, mapping discrete​‌ labels into continuous data​​ introduces unavoidable shortcomings, complicating​​​‌ the evaluation of range-based​ anomalies. Notably, the choice​‌ of evaluation measure may​​ significantly bias the experimental​​​‌ outcome. Despite over six​ decades of attention, there​‌ has never been a​​ large-scale systematic quantitative and​​​‌ qualitative analysis of time-series​ AD evaluation measures. This​‌ paper extensively evaluates quality​​ measures for time-series AD​​​‌ to assess their robustness​ under noise, misalignments, and​‌ different anomaly cardinality ratios.​​ Our results indicate that​​​‌ measures producing quality values​ independently of a threshold​‌ (i.e., AUC-ROC and AUC-PR)​​ are more suitable for​​​‌ time-series AD. Motivated by​ this observation, we first​‌ extend the AUC-based measures​​ to account for range-based​​​‌ anomalies. Then, we introduce​ a new family of​‌ parameter-free and threshold-independent measures,​​ VUS (Volume Under the​​​‌ Surface), to evaluate methods​ while varying parameters. Our​‌ findings demonstrate that our​​ four measures are significantly​​​‌ more robust in assessing​ the quality of time-series​‌ AD methods.
• Functional Description:​​
  The receiver operator characteristic​​​‌ (ROC) curve and the​ area under the curve​‌ (AUC) are widely used​​ to compare the performance​​​‌ of different anomaly detectors.​ They mainly focus on​‌ point-based detection. However, the​​ detection of collective anomalies​​​‌ concerns two factors: whether​ this outlier is detected​‌ and what percentage of​​ this outlier is detected.​​​‌ The first factor is​ not reflected in the​‌ AUC. Another problem is​​ the possible shift between​​​‌ the anomaly score and​ the real outlier due​‌ to the application of​​ the sliding window. To​​​‌ tackle these problems, we​ incorporate the idea of​‌ range-based precision and recall,​​ and suggest the range-based​​​‌ ROC and its counterpart​ in the precision-recall space,​‌ which provides a new​​ evaluation for the collective​​​‌ anomalies. We finally introduce​ a new measure VUS​‌ (Volume Under the Surface)​​ which corresponds to the​​​‌ averaged range-based measure when​ we vary the range​‌ size. We demonstrate in​​ a large experimental evaluation​​​‌ that the proposed measures​ are significantly more robust​‌ to important criteria (such​​ as lag and noise)​​​‌ and also significantly more​ useful to separate correctly​‌ the accurate from the​​ the inaccurate methods.
• News​​ of the Year:

  We​​​‌ recently published in 2025‌ a new paper introducing‌​‌ two optimized implementations of​​ VUS that significantly reduce​​​‌ the execution time of‌ the initial implementation.

  Publication:‌​‌ https://inria.hal.science/hal-05076186

• URL:
  https://­github.­com/­TheDatumOrg/­VUS
• Publication:​​
  hal-05076186
• Contact:
  Paul Boniol​​​‌
• Participants:
  Paul Boniol, Emmanouil‌ Sylligardos, 9 anonymous participants‌​‌
• Partners:
  Ohio State University,​​ Université Paris-Descartes

6.1.3 TSB-UAD​​​‌

• Keywords:
  Time Series, Anomaly‌ detection, Python, Library
• Scientific‌​‌ Description:
  The detection of​​ anomalies in time series​​​‌ has gained ample academic‌ and industrial attention. However,‌​‌ no comprehensive benchmark exists​​ to evaluate time-series anomaly​​​‌ detection methods. It is‌ common to use (i)‌​‌ proprietary or synthetic data,​​ often biased to support​​​‌ particular claims, or (ii)‌ a limited collection of‌​‌ publicly available datasets. Consequently,​​ we often observe methods​​​‌ performing exceptionally well in‌ one dataset but surprisingly‌​‌ poorly in another, creating​​ an illusion of progress.​​​‌ To address the issues‌ above, we thoroughly studied‌​‌ over one hundred papers​​ to identify, collect, process,​​​‌ and systematically format datasets‌ proposed in the past‌​‌ decades. We summarize our​​ effort in TSB-UAD, a​​​‌ new benchmark to ease‌ the evaluation of univariate‌​‌ time-series anomaly detection methods.​​ Overall, TSB-UAD contains 13766​​​‌ time series with labeled‌ anomalies spanning different domains‌​‌ with high variability of​​ anomaly types, ratios, and​​​‌ sizes. TSB-UAD includes 18‌ previously proposed datasets containing‌​‌ 1980 time series and​​ we contribute two collections​​​‌ of datasets. Specifically, we‌ generate 958 time series‌​‌ using a principled methodology​​ for transforming 126 time-series​​​‌ classification datasets into time‌ series with labeled anomalies.‌​‌ In addition, we present​​ data transformations with which​​​‌ we introduce new anomalies,‌ resulting in 10828 time‌​‌ series with varying complexity​​ for anomaly detection. Finally,​​​‌ we evaluate 12 representative‌ methods demonstrating that TSB-UAD‌​‌ is a robust resource​​ for assessing anomaly detection​​​‌ methods. TSB-UAD provides a‌ valuable, reproducible, and frequently‌​‌ updated resource to establish​​ a leaderboard of univariate​​​‌ time-series anomaly detection methods.‌
• Functional Description:
  TSB-UAD is‌​‌ a new open, end-to-end​​ benchmark suite to ease​​​‌ the evaluation of univariate‌ time-series anomaly detection methods.‌​‌ Overall, TSB-UAD contains 12686​​ time series with labeled​​​‌ anomalies spanning different domains‌ with high variability of‌​‌ anomaly types, ratios, and​​ sizes. Specifically, TSB-UAD includes​​​‌ 18 previously proposed datasets‌ containing 1980 time series‌​‌ from real-world data science​​ applications. Motivated by flaws​​​‌ in certain datasets and‌ evaluation strategies in the‌​‌ literature, we study anomaly​​ types and data transformations​​​‌ to contribute two collections‌ of datasets. Specifically, we‌​‌ generate 958 time series​​ using a principled methodology​​​‌ for transforming 126 time-series‌ classification datasets into time‌​‌ series with labeled anomalies.​​ In addition, we present​​​‌ a set of data‌ transformations with which we‌​‌ introduce new anomalies in​​ the public datasets, resulting​​​‌ in 10828 time series‌ (92 datasets) with varying‌​‌ difficulty for anomaly detection.​​
• URL:
  https://­tsb-uad.­readthedocs.­io/­en/­latest/
• Contact:
  Paul​​​‌ Boniol
• Participants:
  Paul Boniol,‌ Emmanouil Sylligardos, 5 anonymous‌​‌ participants
• Partners:
  Université Paris-Descartes,​​ Ohio State University

6.1.4​​​‌ ADecimo

• Name:
  A Web-app‌ for the Evaluation of‌​‌ Model selection for Anomaly​​ Detection in Time Series​​​‌
• Keywords:
  Time Series, Anomaly‌ detection, Web Application
• Scientific‌​‌ Description:
  Anomaly detection is​​​‌ a fundamental task for​ time-series analytics with important​‌ implications for the downstream​​ performance of many applications.​​​‌ Despite increasing academic interest​ and the large number​‌ of methods proposed in​​ the literature, recent benchmark​​​‌ and evaluation studies demonstrated​ that there exists no​‌ single best anomaly detection​​ method when applied to​​​‌ heterogeneous time series datasets.​ Therefore, the only scalable​‌ and viable solution to​​ solve anomaly detection over​​​‌ very different time series​ collected from diverse domains​‌ is to propose a​​ model selection method that​​​‌ will choose, based on​ time series characteristics, the​‌ best anomaly detection method​​ to run. This paper​​​‌ describes ADecimo, a modular​ and extensible web application​‌ that helps users understand​​ the performance of time​​​‌ series classification algorithms used​ as model selection methods​‌ for time series anomaly​​ detection. Overall, our system​​​‌ enables users to compare​ 17 different classifiers over​‌ 1980 time series, and​​ decide on the most​​​‌ suitable time series classification​ method for their own​‌ time series and use​​ cases.
• Functional Description:
  We​​​‌ present here ADecimo, a​ modular and extensible web​‌ application that helps users​​ understand the performance of​​​‌ time series classification algorithms​ used as model selection​‌ methods for time series​​ anomaly detection. Overall, our​​​‌ system enables users to​ compare 17 different classifiers​‌ over 1980 time series,​​ and decide on the​​​‌ most suitable time series​ classification method for their​‌ own time series and​​ use cases.
• URL:
  https://­adecimots.­streamlit.­app/​​​‌
• Publication:
  hal-04590326
• Contact:
  Paul​ Boniol
• Participants:
  Paul Boniol,​‌ Emmanouil Sylligardos, 3 anonymous​​ participants

6.1.5 MSAD

• Name:​​​‌
  Model Selection for Anomaly​ Detection
• Keywords:
  Time Series,​‌ Machine learning, Classification, Ensemble​​ classifier, Python
• Scientific Description:​​​‌
  Anomaly detection is a​ fundamental task for time-series​‌ analytics with important implications​​ for the downstream performance​​​‌ of many applications. Despite​ increasing academic interest and​‌ the large number of​​ methods proposed in the​​​‌ literature, recent benchmark and​ evaluation studies demonstrated that​‌ no overall best anomaly​​ detection methods exist when​​​‌ applied to very heterogeneous​ time series datasets. Therefore,​‌ the only scalable and​​ viable solution to solve​​​‌ anomaly detection over very​ different time series collected​‌ from diverse domains is​​ to propose a model​​​‌ selection method that will​ select, based on time​‌ series characteristics, the best​​ anomaly detection method to​​​‌ run. Existing AutoML solutions​ are, unfortunately, not directly​‌ applicable to time series​​ anomaly detection, and no​​​‌ evaluation of time series-based​ approaches for model selection​‌ exists. Towards that direction,​​ this paper studies the​​​‌ performance of time series​ classification methods used as​‌ model selection for anomaly​​ detection. Overall, we compare​​​‌ 17 different classifiers over​ 1800 time series, and​‌ we propose the first​​ extensive experimental evaluation of​​​‌ time series classification as​ model selection for anomaly​‌ detection. Our results demonstrate​​ that model selection methods​​​‌ outperform every single anomaly​ detection method while being​‌ in the same order​​ of magnitude regarding execution​​​‌ time. This evaluation is​ the first step to​‌ demonstrate the accuracy and​​ efficiency of time series​​​‌ classification algorithms for anomaly​ detection, and represents a​‌ strong baseline that can​​ then be used to​​ guide the model selection​​​‌ step in general AutoML‌ pipelines.
• Functional Description:
  MSAD‌​‌ proposes a pipeline for​​ model selection based on​​​‌ time series classification and‌ an extensive experimental evaluation‌​‌ of existing classification algorithms​​ for this new pipeline.​​​‌ Our results demonstrate that‌ model selection methods outperform‌​‌ every single anomaly detection​​ method while being in​​​‌ the same order of‌ magnitude regarding execution time.‌​‌
• News of the Year:​​

  In 2025, we published​​​‌ a new paper that‌ extended the model selection‌​‌ pipeline, improving performance in​​ Out-of-Distribution (OoD) settings.

  Paper:​​​‌ https://inria.hal.science/hal-05343228

• URL:
  https://­github.­com/­boniolp/­MSAD
• Publication:‌
  hal-05343228
• Contact:
  Paul Boniol‌​‌
• Participants:
  Emmanouil Sylligardos, Paul​​ Boniol, Pierre Senellart, 2​​​‌ anonymous participants
• Partners:
  Ohio‌ State University, Université Paris-Descartes‌​‌

6.1.6 apxproof

• Keyword:
  LaTeX​​
• Functional Description:
  apxproof is​​​‌ a LaTeX package facilitating‌ the typesetting of research‌​‌ articles with proofs in​​ appendix, a common practice​​​‌ in database theory and‌ theoretical computer science in‌​‌ general. The appendix material​​ is written in the​​​‌ LaTeX code along with‌ the main text which‌​‌ it naturally complements, and​​ it is automatically deferred.​​​‌ The package can automatically‌ send proofs to the‌​‌ appendix, can repeat in​​ the appendix the theorem​​​‌ environments stated in the‌ main text, can section‌​‌ the appendix automatically based​​ on the sectioning of​​​‌ the main text, and‌ supports a separate bibliography‌​‌ for the appendix material.​​
• Release Contributions:
  Fix forward​​​‌ linking when used in‌ conjunction with aliascnt (e.g.,‌​‌ in Springer classes), Compatibility​​ with recent versions of​​​‌ acmart.cls
• News of the‌ Year:
  - Fix forward‌​‌ linking when used in​​ conjunction with aliascnt (e.g.,​​​‌ in Springer classes) -‌ Compatibility with recent versions‌​‌ of acmart.cls - Support​​ for user-defined claimproof environments​​​‌ - Remove forward linking‌ command from PDF bookmarks‌​‌
• URL:
  https://­github.­com/­PierreSenellart/­apxproof
• Contact:
  Pierre​​ Senellart
• Participant:
  Pierre Senellart​​​‌

Knowledge representation and‌ knowledge bases

In 28‌​‌, 27, we​​ establish a correspondence between​​​‌ (fragments of) ${\mathrm{𝒯ℰℒ}}^{◯‌}$, a temporal extension‌​‌ of the $\mathrm{ℰℒ}$ description​​ logic with the LTL​​​‌ operator ${◯}^k$,‌ and some specific kinds‌​‌ of formal grammars, in​​ particular, conjunctive grammars (context-free​​​‌ grammars equipped with the‌ operation of intersection). This‌​‌ connection implies that ${\mathrm{𝒯ℰℒ}}^{◯}$ does not possess​​​‌ the property of ultimate‌ periodicity of models, and‌​‌ further leads to undecidability​​ of query answering in​​​‌ ${\mathrm{𝒯ℰℒ}}^{◯}$, closing‌ a question left open‌​‌ since the introduction of​​ ${\mathrm{𝒯ℰℒ}}^{◯}$. Moreover,​​​‌ it also allows to‌ establish decidability of query‌​‌ answering for some new​​ interesting fragments of ${\mathrm{𝒯ℰℒ‌}}^{◯}$, and to‌ reuse for this purpose‌​‌ existing tools and algorithms​​ for conjunctive grammars.

Consistent​​​‌ query answering

In 17‌, we consider the‌​‌ dichotomy conjecture for consistent​​ query answering under primary​​​‌ key constraints. It states‌ that, for every fixed‌​‌ Boolean conjunctive query q,​​ testing whether q is​​​‌ certain (i.e. whether it‌ evaluates to true over‌​‌ all repairs of a​​​‌ given inconsistent database) is​ either polynomial time or​‌ coNP-complete. This conjecture has​​ been verified for self-join-free​​​‌ and path queries. We​ propose a simple inflationary​‌ fixpoint algorithm for consistent​​ query answering which, for​​​‌ a given database, naively​ computes a set $\mathrm{\Delta ‌}$ of subsets of facts​​ of the database of​​​‌ size at most k,​ where k is the​‌ size of the query​​ q. The algorithm runs​​​‌ in polynomial time and​ can be formally defined​‌ as: (1) Initialize $\mathrm{\Delta }$ with all sets $\mathrm{S‌}$ of at most $\mathrm{k}$ facts such that $\mathrm{S‌}\vDash q$. (2)​​ Add any set $\mathrm{S‌}$ of at most k​ facts to $\Delta$ if​‌ there exists a block​​ $B$ (i.e., a maximal​​​‌ set of facts sharing​ the same key) such​‌ that for every fact​​ $a\in B$ there​​​‌ is a set ${\mathrm{S}}^{\text{'}}\subseteq S\cup ‌\left\{a\right\}$ such​​ that $S^{\text{'}}\in ‌\Delta$. For an​ input database $D$,​‌ the algorithm answers "q​​ is certain" iff $\mathrm{\Delta ‌}$ eventually contains the empty​ set. The algorithm correctly​‌ computes certainty when the​​ query q falls in​​​‌ the polynomial time cases​ of the known dichotomies​‌ for self-join-free queries and​​ path queries. For arbitrary​​​‌ Boolean conjunctive queries, the​ algorithm is an under-approximation:​‌ the query is guaranteed​​ to be certain if​​​‌ the algorithm claims so.​ However, there are polynomial​‌ time certain queries (with​​ self-joins) which are not​​​‌ identified as such by​ the algorithm.

The Chase​‌ and Existential Rules

29​​ The chase is a​​​‌ fundamental algorithm with ubiquitous​ uses in database theory.​‌ Given a database and​​ a set of existential​​​‌ rules (aka tuple-generating dependencies),​ it iteratively extends the​‌ database to ensure that​​ the rules are satisfied​​​‌ in a most general​ way. This process may​‌ not terminate, and a​​ major problem is to​​​‌ decide whether it does.​ This problem has been​‌ studied for a large​​ number of chase variants,​​​‌ which differ by the​ conditions under which a​‌ rule is applied to​​ extend the database. Surprisingly,​​​‌ the complexity of the​ universal termination of the​‌ restricted (aka standard) chase​​ is not fully understood.​​​‌ We close this gap​ by placing universal restricted​‌ chase termination in the​​ analytical hierarchy. This higher​​​‌ hardness is due to​ the fairness condition, and​‌ we propose an alternative​​ condition to reduce the​​​‌ hardness of universal termination.​

In 34, we​‌ address one of the​​ fundamental open questions in​​​‌ the realm of existential​ rules: the conjecture on​‌ the finite controllability of​​ bounded derivation depth rule​​​‌ sets ($\mathrm{𝖻𝖽𝖽}\Rightarrow \mathrm{𝖿𝖼}$). We take​‌ a step toward a​​ positive resolution of this​​​‌ conjecture by demonstrating that​ universal models generated by​‌ bdd rule sets cannot​​ contain arbitrarily large tournaments​​​‌ (arbitrarily directed cliques) without​ entailing a loop query,​‌ $\exists xE(x,x)‌$. This simple yet​ elegant result narrows the​‌ space of potential counterexamples​​ to the ($\mathrm{𝖻𝖽𝖽‌}\Rightarrow \mathrm{𝖿𝖼}$) conjecture.​

Other aspects of theoretical​‌ computer science

Our research​​ occasionally touches other aspects​​ of theoretical computer science​​​‌ not related to data‌ management.

In 31,‌​‌ we introduce an extension​​ of fixed-point logic (FP)​​​‌ with a group-order operator‌ (ord), that computes the‌​‌ size of a group​​ generated by a definable​​​‌ set of permutations. This‌ operation is a generalization‌​‌ of the rank operator​​ (rk). We show that​​​‌ FP + ord constitutes‌ a new candidate logic‌​‌ for the class of​​ polynomial-time computable queries (P).​​​‌ As was the case‌ for FP + rk,‌​‌ the model-checking of FP​​ + ord formulae is​​​‌ polynomial-time computable. Moreover, the‌ query separating FP +‌​‌ rk from P exhibited​​ by Lichter in his​​​‌ recent breakthrough is definable‌ in FP + ord.‌​‌ Precisely, we show that​​ FP + ord canonizes​​​‌ structures with Abelian colors,‌ a class of structures‌​‌ which contains Lichter's counter-example.​​ This proof involves expressing​​​‌ a fragment of the‌ group-theoretic approach to graph‌​‌ canonization in the logic​​ FP + ord.

Inconsistent‌​‌ knowledge bases

Repair-based semantics​​ have been extensively studied​​​‌ as a means of‌ obtaining meaningful answers to‌​‌ queries posed over inconsistent​​ knowledge bases (KBs). While​​​‌ several works have considered‌ how to exploit a‌​‌ priority relation between facts​​ to select optimal repairs,​​​‌ the question of how‌ to specify such preferences‌​‌ remains largely unaddressed. This​​ motivates us in 23​​​‌, 22 to introduce‌ a declarative rule-based framework‌​‌ for specifying and computing​​ a priority relation between​​​‌ conflicting facts. As the‌ expressed preferences may contain‌​‌ undesirable cycles, we consider​​ the problem of determining​​​‌ when a set of‌ preference rules always yields‌​‌ an acyclic relation, and​​ we also explore a​​​‌ pragmatic approach that extracts‌ an acyclic relation by‌​‌ applying various cycle removal​​ techniques. Towards an end-to-end​​​‌ system for querying inconsistent‌ KBs, we present a‌​‌ preliminary implementation and experimental​​ evaluation of the framework,​​​‌ which employs answer set‌ programming to evaluate the‌​‌ preference rules, apply the​​ desired cycle resolution techniques​​​‌ to obtain a priority‌ relation, and answer queries‌​‌ under prioritized-repair semantics.

In​​ 25, 24,​​​‌ we explore the issue‌ of inconsistency handling in‌​‌ DatalogMTL, an extension of​​ Datalog with metric temporal​​​‌ operators. Since facts are‌ associated with time intervals,‌​‌ there are different manners​​ to restore consistency when​​​‌ they contradict the rules,‌ such as removing facts‌​‌ or modifying their time​​ intervals. Our first contribution​​​‌ is the definition of‌ relevant notions of conflicts‌​‌ (minimal explanations for inconsistency)​​ and repairs (possible ways​​​‌ of restoring consistency) for‌ this setting and the‌​‌ study of the properties​​ of these notions and​​​‌ the associated inconsistency-tolerant semantics.‌ Our second contribution is‌​‌ a data complexity analysis​​ of the tasks of​​​‌ generating a single conflict‌ / repair and query‌​‌ entailment under repair-based semantics.​​

Provenance and probability management​​​‌

Ensemble methods aggregate the‌ predictions of multiple models‌​‌ by some form of​​ weighted voting. In 33​​​‌, we consider the‌ impact of the choice‌​‌ of the assignment of​​​‌ voting power to every​ individual model on the​‌ performance of ensemble methods.​​ We empirically and comparatively​​​‌ evaluate the accuracy and​ running time of the​‌ different power voting ensemble​​ methods using standard classifiers​​​‌ and mainstream classification benchmarks.​ The results show that​‌ power ensemble voting outperforms​​ the equal-power baseline, and​​​‌ that unsupervised learning of​ the voting power can​‌ be competitive with respect​​ to supervised learning; within​​​‌ supervised approaches, learning voting​ power through Shapley values​‌ and regression outperforms simply​​ using accuracy.

The Shapley​​​‌ value provides a principled​ framework for attributing marginal​‌ contributions to players in​​ coalitional games. While its​​​‌ axiomatic fairness guarantees have​ made it a cornerstone​‌ of value distribution in​​ economics and multi-agent systems,​​​‌ recent computational advances have​ extended its applicability to​‌ data-driven domains. 32 bridges​​ game-theoretic foundations with probabilistic​​​‌ reasoning by studying Shapley-like​ scores in stochastic environments.​‌ We prove that the​​ expected Shapley value (EShap)​​​‌ – player's average impact​ in a game with​‌ an independent probabilistic setting​​ – coincides with the​​​‌ Shapley value of the​ game whose utility is​‌ the expected utility of​​ the original game (ShapE).​​​‌ This equality, however, fails​ for other power indices,​‌ such as the Banzhaf​​ index, underscoring the Shapley​​​‌ value's specificity of consistency​ in uncertain settings. We​‌ further identify that for​​ a certain class of​​​‌ coefficients (including normalized Banzhaf​ indices) the equality persists,​‌ broadening the scope of​​ reliable attribution mechanisms.

ProvSQL​​​‌ is a PostgreSQL extension​ implementing provenance management and​‌ probabilistic database features. ProvSQL​​ seamlessly extends relational database​​​‌ functionality to support the​ storage, tracking through derivations​‌ and transformations, and querying​​ of metadata that explain​​​‌ and qualify the data​ and query results. In​‌ 40, ProvSQL is​​ used to implement a​​​‌ content-based image retrieval system.​ A deep learning object​‌ detection model identifies objects​​ of selected classes located​​​‌ within the images of​ a large-scale image data​‌ set. The uncertainty associated​​ with object detection is​​​‌ recorded. ProvSQL's provenance model​ incorporates this uncertainty into​‌ the retrieval process, thus​​ facilitating the generation of​​​‌ accurate and reliable results​ and allowing for decision-making​‌ in scenarios with incomplete​​ or uncertain information. The​​​‌ demonstration illustrates how ProvSQL​ handles query processing, uncertainty​‌ tracking, and probability computation.​​ It highlights the utility​​​‌ of a probabilistic database​ for applications dealing with​‌ uncertain data, compared to​​ traditional threshold-based approaches.

In​​​‌ 39, we further​ enhance ProvSQL by enabling​‌ provenance tracking for update​​ operations (DELETE, INSERT, UPDATE).​​​‌ We illustrate the practical​ utility of update provenance​‌ by implementing a temporal​​ database capable of standard​​​‌ operations, including time travel​ (inspecting past database states),​‌ history tracking (monitoring tuple​​ states over time), and​​​‌ undo (reversing previous updates).​ These features rely on​‌ a provenance formalism based​​ on the union-of-intervals m-semiring.​​​‌ Additionally, we emphasize a​ key advantage of using​‌ semiring-based provenance model: its​​ generality allows the same​​​‌ semiring structure to seamlessly​ support various applications, such​‌ as probabilistic databases, by​​ simply modifying the semiring​​​‌ definition.

Mining​​ time series

Recent advances​​​‌ in data collection technology,‌ accompanied by the ever-rising‌​‌ volume and velocity of​​ streaming data, underscore the​​​‌ vital need for time‌ series analytics. In this‌​‌ regard, time-series anomaly detection​​ has been an important​​​‌ activity, entailing various applications‌ in fields such as‌​‌ cyber security, financial markets,​​ law enforcement, and health​​​‌ care. While traditional literature‌ on anomaly detection is‌​‌ centered on statistical measures,​​ the increasing number of​​​‌ machine learning algorithms in‌ recent years calls for‌​‌ a structured, general characterization​​ of the research methods​​​‌ for time-series anomaly detection.‌ In 36, we‌​‌ present a process-centric taxonomy​​ for time-series anomaly detection​​​‌ methods, systematically categorizing traditional‌ statistical approaches and contemporary‌​‌ machine learning techniques. Beyond​​ this taxonomy, we conduct​​​‌ a meta-analysis of the‌ existing literature to identify‌​‌ broad research trends. Given​​ the absence of a​​​‌ one-size-fits-all anomaly detector, we‌ also introduce emerging trends‌​‌ for time-series anomaly detection.​​ Furthermore, we review commonly​​​‌ used evaluation measures and‌ benchmarks, followed by an‌​‌ analysis of benchmark results​​ to provide insights into​​​‌ the impact of different‌ design choices on model‌​‌ performance. Through these contributions,​​ we aim to provide​​​‌ a holistic perspective on‌ time-series anomaly detection and‌​‌ highlight promising avenues for​​ future investigation.

Anomaly detection​​​‌ is a fundamental task‌ for time series analytics‌​‌ with important implications for​​ the downstream performance of​​​‌ many applications. Despite increasing‌ academic interest and the‌​‌ large number of methods​​ proposed in the literature,​​​‌ recent benchmarks and evaluation‌ studies demonstrated that no‌​‌ overall best anomaly detection​​ methods exist when applied​​​‌ to very heterogeneous time‌ series datasets. Therefore, the‌​‌ only scalable and viable​​ solution to solve anomaly​​​‌ detection over very different‌ time series collected from‌​‌ diverse domains is to​​ propose a model selection​​​‌ method that will select,‌ based on time series‌​‌ characteristics, the best anomaly​​ detection methods to run.​​​‌ Existing AutoML solutions are,‌ unfortunately, not directly applicable‌​‌ to time series anomaly​​ detection, and no evaluation​​​‌ of time series-based approaches‌ for model selection exists.‌​‌ Towards that direction, 19​​ studies the performance of​​​‌ time series classification methods‌ used as model selection‌​‌ for anomaly detection. In​​ total, we evaluate 234​​​‌ model configurations derived from‌ 16 base classifiers across‌​‌ more than 1980 time​​ series, and we propose​​​‌ the first extensive experimental‌ evaluation of time series‌​‌ classification as model selection​​ for anomaly detection. Our​​​‌ results demonstrate that model‌ selection methods outperform every‌​‌ single anomaly detection method​​ while being in the​​​‌ same order of magnitude‌ regarding execution time. This‌​‌ evaluation is the first​​ step to demonstrate the​​​‌ accuracy and efficiency of‌ time series classification algorithms‌​‌ for anomaly detection, and​​ represents a strong.

In​​​‌ contrast to other domains‌ where AD mainly focuses‌​‌ on point-based anomalies (i.e.,​​ outliers in standalone observations),​​​‌ AD for time series‌ is also concerned with‌​‌ range-based anomalies (i.e., outliers​​ spanning multiple observations). Nevertheless,​​​‌ it is common to‌ use traditional point-based information‌​‌ retrieval measures, such as​​​‌ Precision, Recall, and F-score,​ to assess the quality​‌ of methods by thresholding​​ the anomaly score to​​​‌ mark each point as​ an anomaly or not.​‌ However, mapping discrete labels​​ into continuous data introduces​​​‌ unavoidable shortcomings, complicating the​ evaluation of range-based anomalies.​‌ Notably, the choice of​​ evaluation measure may significantly​​​‌ bias the experimental outcome.​ Despite over six decades​‌ of attention, there has​​ never been a large-scale​​​‌ systematic quantitative and qualitative​ analysis of time-series AD​‌ evaluation measures. 15 extensively​​ evaluates quality measures for​​​‌ time-series AD to assess​ their robustness under noise,​‌ misalignments, and different anomaly​​ cardinality ratios. Our results​​​‌ indicate that measures producing​ quality values independently of​‌ a threshold (i.e., AUC-ROC​​ and AUC-PR) are more​​​‌ suitable for time-series AD.​ Motivated by this observation,​‌ we first extend the​​ AUC-based measures to account​​​‌ for range-based anomalies. Then,​ we introduce a new​‌ family of parameter-free and​​ threshold-independent measures, Volume Under​​​‌ the Surface (VUS), to​ evaluate methods while varying​‌ parameters. We also introduce​​ two optimized implementations for​​​‌ VUS that reduce significantly​ the execution time of​‌ the initial implementation. Our​​ findings demonstrate that our​​​‌ four measures are significantly​ more robust in assessing​‌ the quality of time-series​​ AD methods.

Motif Discovery​​​‌ involves identifying recurring patterns​ and locating their occurrences​‌ within a time series​​ without prior knowledge about​​​‌ their shape or location.​ In practice, Motif Discovery​‌ faces several data-related challenges,​​ leading to various definitions​​​‌ of the problem and​ multiple algorithms addressing these​‌ challenges to different extents.​​ However, there has been​​​‌ no systematic evaluation and​ comparison of these diverse​‌ approaches. Consequently, 18 presents​​ a comprehensive literature review​​​‌ covering data-related challenges, motif​ definitions, and algorithms. We​‌ also analyze the strengths​​ and limitations of algorithms​​​‌ carefully chosen to represent​ the literature diversity. The​‌ analysis is structured around​​ key research questions identified​​​‌ from our review. Our​ experimental findings provide practical​‌ guidelines for selecting Motif​​ Discovery algorithms suitable for​​​‌ a given task and​ suggest directions for future​‌ research.

Time series clustering​​ poses a significant challenge​​​‌ with diverse applications across​ domains. A prominent drawback​‌ of existing solutions lies​​ in their limited interpretability,​​​‌ often confined to presenting​ users with centroids. In​‌ addressing this gap, 16​​ presents k-Graph, an unsupervised​​​‌ method explicitly crafted to​ augment interpretability in time​‌ series clustering. Leveraging a​​ graph representation of time​​​‌ series subsequences, k-Graph constructs​ multiple graph representations based​‌ on different subsequence lengths.​​ This feature accommodates variable-length​​​‌ time series without requiring​ users to predetermine subsequence​‌ lengths. Our experimental results​​ reveal that k-Graph outperforms​​​‌ current state-of-the-art time series​ clustering algorithms in accuracy,​‌ while providing users with​​ meaningful explanations and interpretations​​​‌ of the clustering outcomes.​

Time series clustering is​‌ important for identifying patterns​​ in these datasets. However,​​​‌ prevailing methods often encounter​ obstacles in maintaining data​‌ relationships and ensuring interpretability.​​ We present in 26​​​‌ Graphint, an innovative system​ based on the $\mathrm{k‌}$-Graph methodology that addresses​​ these challenges. Graphint integrates​​​‌ a robust time series​ clustering algorithm with an​‌ interactive tool for comparison​​ and interpretation. More precisely,​​ our system allows users​​​‌ to compare results against‌ competing approaches, identify discriminative‌​‌ subsequences within specified datasets,​​ and visualize the critical​​​‌ information utilized by $\mathrm{k‌}$-Graph to generate outputs.‌​‌ Overall, Graphint offers a​​ comprehensive solution for extracting​​​‌ actionable insights from complex‌ temporal datasets.

Time series‌​‌ segmentation is a fundamental​​ task in analyzing temporal​​​‌ data across various domains,‌ from human activity recognition‌​‌ to energy monitoring. While​​ numerous state-of-the-art methods have​​​‌ been developed to tackle‌ this problem, the evaluation‌​‌ of their performance remains​​ critically limited. Existing measures​​​‌ predominantly focus on change‌ point accuracy or rely‌​‌ on point-based measures such​​ as Adjusted Rand Index​​​‌ (ARI), which fail to‌ capture the quality of‌​‌ the detected segments, ignore​​ the nature of errors,​​​‌ and offer limited interpretability.‌ In 30, we‌​‌ address these shortcomings by​​ introducing two novel evaluation​​​‌ measures: WARI (Weighted Adjusted‌ Rand Index), that accounts‌​‌ for the position of​​ segmentation errors, and SMS​​​‌ (State Matching Score), a‌ fine-grained measure that identifies‌​‌ and scores four fundamental​​ types of segmentation errors​​​‌ while allowing error-specific weighting.‌ We empirically validate WARI‌​‌ and SMS on synthetic​​ and real-world benchmarks, showing​​​‌ that they not only‌ provide a more accurate‌​‌ assessment of segmentation quality​​ but also uncover insights,​​​‌ such as error provenance‌ and type, that are‌​‌ inaccessible with traditional measures.​​

In recent years, electricity​​​‌ suppliers have installed millions‌ of smart meters worldwide‌​‌ to improve the management​​ of the smart grid​​​‌ system. These meters collect‌ a large amount of‌​‌ electrical consumption data to​​ produce valuable information to​​​‌ help consumers reduce their‌ electricity footprint. However, having‌​‌ non-expert users (e.g., consumers​​ or sales advisors) understand​​​‌ these data and derive‌ usage patterns for different‌​‌ appliances has become a​​ significant challenge for electricity​​​‌ suppliers because these data‌ record the aggregated behavior‌​‌ of all appliances. At​​ the same time, ground-truth​​​‌ labels (which could train‌ appliance detection and localization‌​‌ models) are expensive to​​ collect and extremely scarce​​​‌ in practice. 37 introduces‌ DeviceScope, an interactive tool‌​‌ designed to facilitate understanding​​ smart meter data by​​​‌ detecting and localizing individual‌ appliance patterns within a‌​‌ given time period. Our​​ system is based on​​​‌ CamAL (Class Activation Map-based‌ Appliance Localization), a novel‌​‌ weakly supervised approach for​​ appliance localization that only​​​‌ requires the knowledge of‌ the existence of an‌​‌ appliance in a household​​ to be trained.

Improving​​​‌ smart grid system management‌ is crucial in the‌​‌ fight against climate change,​​ and enabling consumers to​​​‌ play an active role‌ in this effort is‌​‌ a significant challenge for​​ electricity suppliers. In this​​​‌ regard, millions of smart‌ meters have been deployed‌​‌ worldwide in the last​​ decade, recording the main​​​‌ electricity power consumed in‌ individual households. This data‌​‌ produces valuable information that​​ can help them reduce​​​‌ their electricity footprint; nevertheless,‌ the collected signal aggregates‌​‌ the consumption of the​​ different appliances running simultaneously​​​‌ in the house, making‌ it difficult to apprehend.‌​‌ Non-Intrusive Load Monitoring (NILM)​​ refers to the challenge​​​‌ of estimating the power‌ consumption, pattern, or on/off‌​‌ state activation of individual​​​‌ appliances using the main​ smart meter signal. Recent​‌ methods proposed to tackle​​ this task are based​​​‌ on a fully supervised​ deep-learning approach that requires​‌ both the aggregate signal​​ and the ground truth​​​‌ of individual appliance power.​ However, such labels are​‌ expensive to collect and​​ extremely scarce in practice,​​​‌ as they require conducting​ intrusive surveys in households​‌ to monitor each appliance.​​ In 38, we​​​‌ introduce CamAL, a weakly​ supervised approach for appliance​‌ pattern localization that only​​ requires information on the​​​‌ presence of an appliance​ in a household to​‌ be trained. CamAL merges​​ an ensemble of deep-learning​​​‌ classifiers combined with an​ explainable classification method to​‌ be able to localize​​ appliance patterns. Our experimental​​​‌ evaluation, conducted on 4​ real-world datasets, demonstrates that​‌ CamAL significantly outperforms existing​​ weakly supervised baselines and​​​‌ that current SotA fully​ supervised NILM approaches require​‌ significantly more labels to​​ reach CamAL performances.

Information​​​‌ Extraction

35, which​ is situated within the​‌ TheoremKB 41 project, presents​​ TheoremView, a novel framework​​​‌ for extracting proofs and​ theorems from raw PDF​‌ scientific papers without requiring​​ LaTeX source files. Our​​​‌ approach combines three modalities​ (font, text, and vision)​‌ with sequential modeling to​​ capture long-term dependencies and​​​‌ layout information. By eliminating​ OCR preprocessing, TheoremView reduces​‌ computational overhead for real-time​​ applications while providing robust​​​‌ automated theorem extraction.

Graphs,​ and notably RDF graphs,​‌ are a prominent way​​ of sharing data. As​​​‌ data usage democratizes, users​ need help figuring out​‌ the useful content of​​ a graph dataset. In​​​‌ particular, journalists with whom​ we collaborate are interested​‌ in identifying, in a​​ graph, the connections between​​​‌ entities, e.g., people, organizations,​ emails, etc. In 14​‌, we present a​​ novel method for exploring​​​‌ data graphs through their​ data paths connecting Named​‌ Entities (NEs, in short);​​ each data path leads​​​‌ to a tabular-looking set​ of results. NEs are​‌ extracted from the data​​ through dedicated Information Extraction​​​‌ modules. Our method builds​ upon the pre-existing ConnectionLens​‌ platform and follow-up work​​ in the Abstra project,​​​‌ which builds simple, visual​ ER-style summaries of semi-structured​‌ data. The contribution of​​ the present work, and​​​‌ its novelty, is twofold.​ First, we propose a​‌ novel analysis of entity-to-entity​​ paths contained in datasets​​​‌ of any nature, and​ propose a new method​‌ for ranking paths, leveraging​​ a novel Information Extraction​​​‌ module we built on​ top of ChatGPT. Second,​‌ we present an efficient​​ approach to enumerate and​​​‌ compute NE paths, based​ on an algorithm which​‌ automatically recommends sub-paths to​​ materialize, and rewrites the​​​‌ path queries using these​ subpaths. Our experiments demonstrate​‌ the interest of NE​​ paths and the efficiency​​​‌ of our method for​ computing and ranking them.​‌

9.1.1 Participation in​​ other International Programs

• DesCartes​​​‌

  Participants: Pierre Senellart.‌

  • Title:
    Intelligent Modelling for‌​‌ Decision-making in Critical Urban​​ Systems
  • Partner Institution(s):
    CNRS@CREATE,​​​‌ National University of Singapore‌
  • Duration:
    2021–2026
  • Additional info:‌​‌
    DesCartes is a project​​ managed by CNRS@CREATE, a​​​‌ CNRS subsidiary in Singapore‌ and funded by Singapore’s‌​‌ National Research Foundation, with​​ 50 million total budget.​​​‌ Pierre Senellart is involved‌ in the project as‌​‌ one of the French​​ PIs, and became in​​​‌ 2025 Lead PI for‌ one of the workpackages.‌​‌
• International ANR project EQUUS​​

  Participants: Luc Segoufin.​​​‌

  • Title:
    Efficient query answering‌ under updates
  • Partner Institution(s):‌​‌
    TU Ilmenau, Uni. Bayreuth,​​ HU Berlin, CNRS
  • Duration:​​​‌
    2020–2025

9.2.1 Visits of‌​‌ international scientists

9.2.2 Visits to international‌​‌ teams

9.3.1 ANR

• PRC EXPAND​​​‌ (coordinator)

  Participants: Michael Thomazo‌, Camille Bourgaux.‌​‌

  • Title:
    Expanding the reach​​ of ontology-based data access:​​​‌ EXpressivity, exPlanation, and Algorithms‌
  • Partner Institution(s):
    LIRMM, LaBRI,‌​‌ LIMOS, Inria Lille (SPIRALS​​ & D-DAL), IRISA
  • Duration:​​​‌
    2025–2030
  • Budget for Valda:‌
    55 k€ (Inria budget)‌​‌
• PR[AI]RIE-PSAI AI Cluster

  Participants:​​ Pierre Senellart, Camille​​​‌ Bourgaux.

  • Title:
    Paris‌ Artificial Intelligence Research Institute‌​‌ – Paris School of​​ AI
  • Duration:
    2025–2029
  • Funding​​​‌ for Valda:
    575 k€‌ (ENS budget)
• Megyn's Bienvenu‌​‌ INTENDED Chair in Artificial​​ Intelligence

  Participants: Camille Bourgaux​​​‌.

  • Title:
    Intelligent handling‌ of imperfect data
  • Partner‌​‌ Institution(s):
    LaBRI
  • Duration:
    2020–2026​​

9.3.2 Others

• France 2030​​​‌ i-Demo Cyberté project

  Participants:‌ Paul Boniol.

  • Partner‌​‌ institution(s):
    Scality, Inria Rennes​​ (CIDRE)
  • Duration:
    2025–2030
  • Funding​​​‌ for Valda:
    499 k€‌ (Inria budget)
• CNRS MITI‌​‌ nanoNet project

  Participants: Paul​​ Boniol.

  • Title:
    Méthodologie​​​‌ avancée pour la détection‌ des nanoparticules dans les‌​‌ séries temporelles spICP-ToF-MS
  • Partner​​ Institution(s):
    IPGP
  • Duration:
    2025–2026​​​‌

10.1.1 Scientific‌​‌ events: organisation

10.1.2 Scientific​​ events: selection

10.1.3 Journal​​​‌

10.1.4​‌ Invited talks

• Paul Boniol​​ , Anomaly Detection in​​​‌ Time Series: Overview and​ New Trends, Invited​‌ speaker at Orange Innovation​​
• Paul Boniol , An​​​‌ introduction to Time series​ anomaly detection (a data-driven​‌ perspective), Invited speaker​​ for SIDOS at EGC​​​‌ 2025, Strasbourg, France (January​ 2025)
• Pierre Senellart ,​‌ Quels horizons de pratique​​ pour la recherche en​​​‌ IA?, Invited speaker​ at Printemps Couperin, Paris,​‌ France (March 2025)
• Silviu​​ Maniu (Univ. Grenoble–Alpes) &​​​‌ Pierre Senellart , Making​ Provenance and Probabilistic Database​‌ Theory Work in Practice​​, Invited talk at​​​‌ ICDT 2025 (Database Theory​ in Practice), Barcelona, Spain(March​‌ 2025) 21
• Pierre Senellart​​ , Qualitative Evaluation of​​​‌ Academic Careers in Computer​ Science at CNRS. Global​‌ Forum on Development of​​ Computer Science, Tsinghua University,​​​‌ Beijing, Chine, Invited​ keynote speaker at the​‌ Global Forum on Development​​ of Computer Science of​​​‌ Tsinghua University, Beijing, China​ (April 2025) 44
• Pierre​‌ Senellart , Artificial Intelligence.​​ A Personal View.​​​‌ Invited speaker at INSP​ Days, Paris, France (July​‌ 2025)
• Pierre Senellart ,​​ Les BD pourront-elles sauver​​​‌ l'IA?, Panel participant,​ BDA 2025, Toulouse, France​‌ (October 2025)
• Pierre Senellart​​ , Intelligence artificielle: Concepts,​​​‌ modèles et enjeux,​ Invited speaker at Séminaire​‌ scientifique et technique de​​ l'Inrap, Chartres, France (November​​​‌ 2025)
• Paul Boniol ,​ Time Series Anomaly Detection:​‌ The Road to Automatic​​ Solutions, Invited Speaker​​​‌ at the 3rd Macau​ Symposium on Data Science,​‌ Macau SAR, China (December​​ 2025)

10.1.5 Leadership within​​​‌ the scientific community

• Serge​ Abiteboul is a member​‌ of the French Academy​​ of Sciences, of the​​​‌ Academia Europaea, and an​ ACM Fellow.
• Pierre Senellart​‌ was until August 2025​​ is a junior member​​​‌ of the Institut Universitaire​ de France.

10.1.6 Research​‌ administration

• Serge Abiteboul is​​ a member of the​​​‌ scientific committe of the​ Programme Inria Quadrant (PIQ).​‌
• Antoine Gauquier is an​​ elected member of the​​​‌ Conseil d'Administration of ENS-PSL​
• Antoine Gauquier is an​‌ elected member of the​​ DIENS lab council
• Luc​​​‌ Segoufin is a member​ of the Formation Spécialisée​‌ de Site (FSS) of​​ the Inria Paris research​​​‌ centre.
• Pierre Senellart is​ Vice-President of PSL University​‌ in charge of Digital​​ infrastructure and IT convergence.​​​‌ 48
• Pierre Senellart was​ until August 2025 the​‌ president of section 6​​ of the National Committee​​ for Scientific Research. 43​​​‌ As a representative of‌ CoNRS, Pierre Senellart was‌​‌ in the Hcéres evaluation​​ committee of the IRIT​​​‌ research unit, and president‌ of the evaluation committee‌​‌ of the LIRMM research​​ unit.
• Pierre Senellart was​​​‌ until August 2025 a‌ member of the board‌​‌ of the conference of​​ presidents of the national​​​‌ committee (CPCN) and as‌ such a member of‌​‌ the coordination of managing​​ parties of the national​​​‌ committee (C3N).
• Pierre Senellart‌ is deputy director of‌​‌ the DI ENS laboratory,​​ joint between ENS, CNRS,​​​‌ and Inria.
• Pierre Senellart‌ is the scientific resource‌​‌ person for Scientific information​​ & edition of the​​​‌ Inria Paris centre.
• Pierre‌ Senellart is the vice-president‌​‌ of the Gilles Kahn​​ PhD award of Société​​​‌ Informatique de France.
• Pierre‌ Senellart is a member‌​‌ of the strategic orientation​​ committee of ISIMA.​​​‌
• Michael Thomazo is a‌ deputy director of the‌​‌ École Doctorale Sciences Mathématiques​​ de Paris-Centre (ED386)
• We​​​‌ participated in the following‌ hiring committee within universities:‌​‌
  • Camille Bourgaux , Maître​​ de conférences, ENSEIRB-MATMECA-Bordeaux INP​​​‌

10.2.1 Supervision‌​‌

• PhD defended: Anatole Dahan,​​ The Role of Permutation​​​‌ Groups in the Search‌ for a Logic for‌​‌ Polynomial Time, 2020–2025,​​​‌ Arnaud Durand (Université Paris-Cité)​ & Luc Segoufin 42​‌

• PhD in progress: Antoine​​ Gauquier , Intelligent construction​​​‌ of a multimodal and​ heterogeneous data warehouse, with​‌ data traceability, started​​ in September 2023, Pierre​​​‌ Senellart & Ioana Manolescu​ (Inria Cedar)
• PhD in​‌ progress: Lucas Larroque,Extension​​ of rewriting procedures for​​​‌ reasoning using existential rules​, started in September​‌ 2023, Michaël Thomazo
• PhD​​ in progress: Robin Jean​​​‌ , Integration of preferences​ and domain knowledge in​‌ inconsistency-tolerant ontology-based data access​​, started in October​​​‌ 2023, Meghyn Bienvenu (CNRS​ LaBRI) & Camille Bourgaux​‌
• PhD in progress: Aryak​​ Sen , Scalability of​​​‌ a data provenance and​ probability management system,​‌ started in February 2024,​​ Silviu Maniu (Université Grenoble​​​‌ Alpes) & Pierre Senellart​
• PhD in progress: Emmanouil​‌ Sylligardos , Accuracy and​​ execution time trade-off in​​​‌ ensembling and model selection​ for time series analytics​‌, started in February​​ 2024, Paul Boniol PierreSenellart​​​‌
• PhD in progress: Felix​ Chavelli , Graph representations​‌ for multivariate time series​​ analytics, started in​​​‌ October 2024, Paul Boniol​ & Michaël Thomazo
• PhD​‌ in progress: Pratik Karmakar​​ , Quality, uncertainty, and​​​‌ lineage of data,​ Stéphane Bressan (NUS, deceased),​‌ Tan Kian-Lee (NUS), &​​ Pierre Senellart (as he​​​‌ is based in Singapore,​ he is not considered​‌ a Valda member)
• PhD​​ in progress: Marijan Soric​​​‌ , Exploitation et structuration​ des données et des​‌ connaissances géologiques hétérogènes,​​ started in March 2025,​​​‌ Pierre Senellart , Ioana​ Manolescu (Inria Cedar), &​‌ Cécile Gracianne (BRGM)
• PhD​​ in progress: Magali Parrino​​​‌ , Détection non-supervisée d’anomalies​ dans des flux continus​‌ de séries temporelles multivariées​​, started in July​​​‌ 2025, Paul Boniol ,​ Emmanuel Remy (EDF), &​‌ Pierre Senellart

• PhD in​​ progress: Arthur Lombardo; started​​​‌ in October 2025, Pierre​ Senellart , Antoine Amarilli​‌ (Inria D-DAL) & Mikaël​​ Monet (Inria D-DAL) (as​​​‌ he is based in​ Lille, he is considered​‌ a D-DAL member)

• Master's​​ internship: Arushi Goyal; Pierre​​​‌ Senellart
• Master's internship: Marijan​ Soric; Pierre Senellart and​‌ Ioana Manolescu (Inria Cedar)​​ 45
• M1 research project:​​​‌ Jeanne Coschieri; Michael Thomazo​ & David Carral (Inria​‌ Boreal)
• M1 research project:​​ Paul Raphaël; Michael Thomazo​​​‌ & Lucas Larroque

10.2.2​ Juries

• PhD: François Amat​‌ [reviewer], Institut polytechnique de​​ Paris, Pierre Senellart

10.3.1 Specific official​ responsibilities in science outreach​‌ structures

• Serge Abiteboul ,​​ President of the scientific​​​‌ steering committee of ANR​
• Serge Abiteboul , President​‌ of the AFNIC Foundation​​
• Pierre Senellart is a​​​‌ scientific expert advising the​ Scientific and Ethical Committee​‌ of Parcoursup and MonMaster,​​ the platforms for the​​​‌ selection of higher-education students​ at the first-year level​‌ and the Master's level.​​ As such, he contributed​​​‌ to the 7th yearly​ report of the committee​‌ to the French parliament​​

10.3.2 Productions (articles, videos,​​​‌ podcasts, serious games, ...)​

• Serge Abiteboul , editor​‌ of the binaire blog,​​ which moved from the​​​‌ blog platform of Le​ Monde to that of​‌ La Recherche
• Serge Abiteboul​​ , codirector of the​​​‌ Parlez-moi d'IA podcast on​ Cause commune
• Serge Abiteboul​‌ , co-author of articles​​ on theatre and computer​​ science 46, 47​​​‌

10.3.3 Participation in Live‌ events

• Serge Abiteboul ,‌​‌ co-organizer with French Senator​​ Ghislaine Senée of a​​​‌ Colloquium at the Senate‌ : Les données au‌​‌ service des territoires intelligents​​
• Serge Abiteboul , co-organizer​​​‌ with Isabelle Hilali from‌ Datacraft ot eh conference:‌​‌ Quantum & Intelligence artificielle​​ : vers une convergence​​​‌ des ruptures technologiques ?‌

International journals

• 14​​ articleN.Nelly Barret​​​‌, A.Antoine Gauquier​, J.-J.Jia-Jean Law​‌ and I.Ioana Manolescu​​. Finding meaningful paths​​​‌ in heterogeneous graphs with​ PathWays.Information Systems​‌127January 2025,​​ 102463HALDOIback​​​‌ to text
• 15 article​P.Paul Boniol,​‌ A. K.Ashwin K.​​ Krishna, M.Marine​​​‌ Bruel, Q.Qinghua​ Liu, M.Mingyi​‌ Huang, T.Themis​​ Palpanas, R. S.​​​‌Ruey S. Tsay,​ A.Aaron Elmore,​‌ M. J.Michael J.​​ Franklin and J.John​​​‌ Paparrizos. VUS: Effective​ and Efficient Accuracy Measures​‌ for Time-Series Anomaly Detection​​.The VLDB Journal​​​‌3432February 2025​HALDOIback to​‌ text
• 16 articleP.​​Paul Boniol, D.​​​‌Donato Tiano, A.​Angela Bonifati and T.​‌Themis Palpanas. $\mathrm{k}$-Graph: A Graph Embedding​​​‌ for Interpretable Time Series​ Clustering.IEEE Transactions​‌ on Knowledge and Data​​ Engineering2025, 1-14​​​‌HALDOIback to​ text
• 17 articleD.​‌Diego Figueira, A.​​Anantha Padmanabha, L.​​​‌Luc Segoufin and C.​Cristina Sirangelo. A​‌ Simple Algorithm for Consistent​​ Query Answering under Primary​​​‌ Keys.Logical Methods​ in Computer Science21​‌1February 2025HAL​​DOIback to text​​​‌
• 18 articleV.Valerio​ Guerrini, T.Thibaut​‌ Germain, C.Charles​​ Truong, L.Laurent​​​‌ Oudre and P.Paul​ Boniol. Time Series​‌ Motif Discovery: A Comprehensive​​ Evaluation.Proceedings of​​​‌ the VLDB Endowment (PVLDB)​187August 2025​‌, 2226 - 2239​​HALDOIback to​​​‌ text
• 19 articleE.​Emmanouil Sylligardos, J.​‌John Paparrizos, T.​​Themis Palpanas, P.​​​‌Pierre Senellart and P.​Paul Boniol. MSAD:​‌ A Deep Dive into​​ Model Selection for Time​​​‌ series Anomaly Detection.​The VLDB Journal34​‌6October 2025,​​ 72HALDOIback​​​‌ to text

Invited conferences​

• 20 inproceedingsC.Camille​‌ Bourgaux. Inconsistency-Tolerant Semantics​​ Based on (Preferred) Repairs​​​‌.RW 2025 -​ 21st Reasoning Web International​‌ Summer SchoolReasoning Web​​ Summer School (RW 2025)​​Istanbul, TurkeySeptember 2025​​​‌HALDOIback to‌ text
• 21 inproceedingsS.‌​‌Silviu Maniu and P.​​Pierre Senellart. Database​​​‌ Theory in Action: Making‌ Provenance and Probabilistic Database‌​‌ Theory Work in Practice​​ (Invited Talk).ICDT​​​‌ 2025 - International Conference‌ on Database TheoryBarcelona,‌​‌ SpainMarch 2025HAL​​DOIback to text​​​‌

International peer-reviewed conferences

• 22‌ inproceedingsM.Meghyn Bienvenu‌​‌, C.Camille Bourgaux​​, K.Katsumi Inoue​​​‌ and R.Robin Jean‌. A Rule-Based Approach‌​‌ to Specifying Preferences over​​ Conflicting Facts and Querying​​​‌ Inconsistent Knowledge Bases (Extended‌ Abstract).Proceedings of‌​‌ the 38th International Workshop​​ on Description Logics (DL​​​‌ 2025)DL 2025 -‌ 38th International Workshop on‌​‌ Description LogicsOpole, Poland​​September 2025HALback​​​‌ to text
• 23 inproceedings‌M.Meghyn Bienvenu,‌​‌ C.Camille Bourgaux,​​ K.Katsumi Inoue and​​​‌ R.Robin Jean.‌ A Rule-Based Approach to‌​‌ Specifying Preferences over Conflicting​​ Facts and Querying Inconsistent​​​‌ Knowledge Bases.Proceedings‌ of the 22nd International‌​‌ Conference on Principles of​​ Knowledge Representation and Reasoning​​​‌ (KR 2025)KR 2025‌ - 22nd International Conference‌​‌ on Principles of Knowledge​​ Representation and ReasoningMelbourne,​​​‌ AustraliaNovember 2025HAL‌back to text
• 24‌​‌ inproceedingsM.Meghyn Bienvenu​​, C.Camille Bourgaux​​​‌ and A.Atefe Khodadaditaghanaki‌. Inconsistency Handling in‌​‌ DatalogMTL (Extended Abstract).​​Proceedings of the 38th​​​‌ International Workshop on Description‌ Logics (DL 2025)DL‌​‌ 2025 - 38th International​​ Workshop on Description Logics​​​‌Opole, PolandSeptember 2025‌HALback to text‌​‌
• 25 inproceedingsM.Meghyn​​ Bienvenu, C.Camille​​​‌ Bourgaux and A.Atefe‌ Khodadaditaghanaki. Inconsistency Handling‌​‌ in DatalogMTL.IJCAI​​ 2025 - Thirty-Fourth International​​​‌ Joint Conference on Artificial‌ IntelligenceMontreal, CanadaInternational‌​‌ Joint Conferences on Artificial​​ Intelligence OrganizationAugust 2025​​​‌HALDOIback to‌ text
• 26 inproceedingsP.‌​‌Paul Boniol, D.​​Donato Tiano, A.​​​‌Angela Bonifati and T.‌Themis Palpanas. Graphint:‌​‌ Graph-based Time Series Clustering​​ Visualisation Tool.ICDE​​​‌ 2025 - IEEE 41th‌ International Conference on Data‌​‌ EngineeringHong Kong, Hong​​ Kong SAR ChinaMarch​​​‌ 2025HALback to‌ text
• 27 inproceedingsC.‌​‌Camille Bourgaux, A.​​Anton Gnatenko and M.​​​‌Michaël Thomazo. Analysing‌ Temporal Reasoning in Description‌​‌ Logics Using Formal Grammars​​ (Extended Abstract).Proceedings​​​‌ of the 38th International‌ Workshop on Description Logics‌​‌ (DL 2025)DL 2025​​ - 38th International Workshop​​​‌ on Description LogicsOpole,‌ PolandSeptember 2025HAL‌​‌back to textback​​ to text
• 28 inproceedings​​​‌C.Camille Bourgaux,‌ A.Anton Gnatenko and‌​‌ M.Michaël Thomazo.​​ Analysing Temporal Reasoning in​​​‌ Description Logics Using Formal‌ Grammars.Proceedings of‌​‌ the 28th European Conference​​ on Artificial IntelligenceECAI'25​​​‌ -- 28th European Conference‌ on Artificial IntelligenceBologna,‌​‌ ItalyOctober 2025HAL​​back to textback​​​‌ to text
• 29 inproceedings‌D.David Carral,‌​‌ L.Lukas Gerlach,​​ L.Lucas Larroque and​​​‌ M.Michaël Thomazo.‌ Restricted Chase Termination: You‌​‌ Want More than Fairness​​.ACM digital library​​​‌PODS 2025 - ACM‌ SIGMOD/PODS International Conference on‌​‌ Management of Data3​​​‌Proceedings of the ACM​ on management of data​‌2Berlin, GermanyJune​​ 2025, 1-17HAL​​​‌DOIback to text​
• 30 inproceedingsF.Félix​‌ Chavelli, P.Paul​​ Boniol and M.Michaël​​​‌ Thomazo. Toward Interpretable​ Evaluation Measures for Time​‌ Series Segmentation.NeurIPS​​ 2025 - 39th Annual​​​‌ Conference on Neural Information​ Processing SystemsSan Diego,​‌ United StatesarXiv2025​​HALback to text​​​‌
• 31 inproceedingsA.Anatole​ Dahan. Group Order​‌ Logic.LICS 2025​​ - Logic in Computer​​​‌ ScienceSingapore, SingaporearXiv​May 2025HALDOI​‌back to text
• 32​​ inproceedingsP.Pratik Karmakar​​​‌, A.Antoine Gauquier​ and P.Pierre Senellart​‌. Expected Shapley Value​​ is Shapley Value for​​​‌ Expected Utility Game.​ECSQARU 2025 - 18th​‌ European Conference on Symbolic​​ and Quantitative Approaches to​​​‌ Reasoning with UncertaintyHagen,​ GermanySeptember 2025HAL​‌back to text
• 33​​ inproceedingsP.Pratik Karmakar​​​‌, A.Angelo Saadeh​, P.Pierre Senellart​‌ and S.Stéphane Bressan​​. Discovering Voting Power​​​‌ for Ensemble Methods.​DEXA 2025 - International​‌ Conference on Database and​​ Expert Systems ApplicationsBangkok,​​​‌ ThailandAugust 2025HAL​back to text
• 34​‌ inproceedingsL.Lucas Larroque​​, P.Piotr Ostropolski-Nalewaja​​​‌ and M.Michaël Thomazo​. No Cliques Allowed:​‌ The Next Step Towards​​ BDD/FC Conjecture.Proceedings​​​‌ of the ACM on​ Management of DataPODS​‌ 2025 - ACM SIGMOD/PODS​​ International Conference on Management​​​‌ of Data3Berlin,​ GermanyJune 2025,​‌ 1 - 20HAL​​DOIback to text​​​‌
• 35 inproceedingsS.Shrey​ Mishra, N.Neil​‌ Sharma, A.Antoine​​ Gauquier and P.Pierre​​​‌ Senellart. TheoremView: A​ Framework for Extracting Theorem-Like​‌ Environments from Raw PDFs​​.ECIR 2025 -​​​‌ European Conference on Information​ RetrievalLucca, ItalySpringer​‌April 2025, 6​​HALDOIback to​​​‌ text
• 36 inproceedingsJ.​John Paparrizos, P.​‌Paul Boniol, Q.​​Qinghua Liu and T.​​​‌Themis Palpanas. Advances​ in Time-Series Anomaly Detection:​‌ Algorithms, Benchmarks, and Evaluation​​ Measures.KDD '25:​​​‌ The 31st ACM SIGKDD​ Conference on Knowledge Discovery​‌ and Data MiningToronto​​ (ON), CanadaACMAugust​​​‌ 2025, 6151 -​ 6161HALDOIback​‌ to text
• 37 inproceedings​​A.Adrien Petralia,​​​‌ P.Paul Boniol,​ P.Philippe Charpentier and​‌ T.Themis Palpanas.​​ DeviceScope: An Interactive App​​​‌ to Detect and Localize​ Appliance Patterns in Electricity​‌ Consumption Time Series.​​ICDE 2025 - IEEE​​​‌ 41th International Conference on​ Data EngineeringHong Kong,​‌ Hong Kong SAR China​​May 2025HALback​​​‌ to text
• 38 inproceedings​A.Adrien Petralia,​‌ P.Paul Boniol,​​ P.Philippe Charpentier and​​​‌ T.Themis Palpanas.​ Few Labels are all​‌ you need: A Weakly​​ Supervised Framework for Appliance​​​‌ Localization in Smart-Meter Series​.ICDE 2025 -​‌ IEEE 41th International Conference​​ on Data EngineeringHong​​​‌ Kong, Hong Kong SAR​ ChinaMay 2025HAL​‌back to text
• 39​​ inproceedingsA. A.Albert​​​‌ Ariel Widiaatmaja, B.​Belkis Djeffal, A.​‌Ashish Dandekar and P.​​Pierre Senellart. Demonstration​​ of ProvSQL Update Provenance​​​‌ through Temporal Databases.‌PW25 - ProvenanceWeekBerlin,‌​‌ GermanyJune 2025HAL​​DOIback to text​​​‌
• 40 inproceedingsF.Fajrian‌ Yunus, P.Pratik‌​‌ Karmakar, P.Pierre​​ Senellart, T.Talel​​​‌ Abdessalem and S.Stéphane‌ Bressan. Using a‌​‌ Probabilistic Database in an​​ Image Retrieval Application.​​​‌EDBT 2025 - 28th‌ International Conference on Extending‌​‌ Database TechnologyBarcelona, Spain​​March 2025HALback​​​‌ to text

National peer-reviewed‌ Conferences

• 41 inproceedingsS.‌​‌Shrey Mishra, A.​​Antoine Gauquier and P.​​​‌Pierre Senellart. Apprentissage‌ multimodal modulaire pour l'extraction‌​‌ de théorèmes et de​​ preuves dans des documents​​​‌ scientifiques longs.Revue‌ des Nouvelles Technologies de‌​‌ l'InformationExtraction et Gestion​​ des Connaissances, EGC'2025Strasbourg,​​​‌ FranceJanuary 2025HAL‌back to text

Doctoral‌​‌ dissertations and habilitation theses​​

• 42 thesisA.Anatole​​​‌ Dahan. The Role‌ of Permutation Groups in‌​‌ the Search for a​​ Logic for Polynomial Time​​​‌.Université Paris Cité‌ (UPC)July 2025HAL‌​‌back to text

Reports​​ & preprints

• 43 report​​​‌N.Nathalie Appel,‌ J.Jérémie Bourdon,‌​‌ N.Nicolas Bousquet,​​ J.Johanne Cohen,​​​‌ A.Antoine Genitrini,‌ P.Patricia Georgeon,‌​‌ Y.Yves Grandvalet,​​ K.Katia Jaffrès-Runser,​​​‌ A.Arnaud Legrand,‌ D.Damian Markham,‌​‌ A.Anca Muscholl,​​ A.Anastasia Paparrizou,​​​‌ L.Loïc Paulevé,‌ M.Michael Poss,‌​‌ M.Maria Gradinariu Potop-Butucaru​​, J.-F.Jean-Florent Raymond​​​‌, R.Romain Rouvoy‌, Y.Yolande Sallent‌​‌, P.Pierre Senellart​​, T.Thomas Seiller​​​‌, Y.-Q.Ye-Qiong Song‌, A.Alain Tchana‌​‌ and H.Hélène Waeselynck​​. Section 06 Sciences​​​‌ de l’information : fondements‌ de l’informatique, calculs, algorithmes,‌​‌ représentations, exploitations: Rapport de​​ conjoncture 2024.CNRS​​​‌2025, 1-20HAL‌back to text

Other‌​‌ scientific publications

• 44 article​​P.Pierre Senellart.​​​‌ Qualitative Evaluation of Academic‌ Researchers in Computer Science:‌​‌ Practices and Reflections from​​ CNRS.Computer Education​​​‌123722025,‌ 36-41HALDOIback‌​‌ to text
• 45 thesis​​M.Marijan Soric.​​​‌ Understanding and Extracting Table‌ Information from BRGM Documents‌​‌.Ecole centrale de​​ LyonFebruary 2025HAL​​​‌back to text

Scientific‌ popularization

• 46 articleR.‌​‌Rémi Ronfard, C.​​Charlotte Truchet and S.​​​‌Serge Abiteboul. Informatique‌ théâtrale.BinaireJanuary‌​‌ 2025, 1-12HAL​​back to text
• 47​​​‌ articleR.Rémi Ronfard‌, C.Charlotte Truchet‌​‌ and S.Serge Abiteboul​​. Régie, captation, mise​​​‌ en scène… Quand l’informatique‌ s’invite au théâtre.‌​‌The Conversation FranceJanuary​​ 2025HALback to​​​‌ text
• 48 articleP.‌Pierre Senellart and N.‌​‌Nathalie Vieira. OnePSL30​​ : transformation.Collection​​​‌ numérique de l’AMUE, Agence‌ de mutualisation des universités‌​‌ et établissements d'enseignement supérieur​​39December 2025HAL​​​‌back to text