Institut des Sciences de l'Information et de leurs Interactions

In economics and perceptual decision-making contextual effects are well documented, where decision weights are adjusted as a function of the distribution of stimuli. Yet, in reinforcement learning literature whether and how contextual information pertaining to decision states is integrated in learning algorithms has received comparably little attention. Here, we investigate reinforcement learning behavior and its computational substrates in a task where we orthogonally manipulate outcome valence and magnitude, resulting in systematic variations in state-values. Model comparison indicates that subjects' behavior is best accounted for by an algorithm which includes both reference point-dependence and range-adaptation-two crucial features of state-dependent valuation. In addition, we find that state-dependent outcome valuation progressively emerges, is favored by increasing outcome information and correlated with explicit understanding of the task structure. Finally, our data clearly show that, while being locally adaptive (for instance in negative valence and small magnitude contexts), state-dependent valuation comes at the cost of seemingly irrational choices, when options are extrapolated out from their original contexts.

It has been over a decade since the first publication of a method dedicated entirely to mapping long-reads. The distinctive characteristics of long reads resulted in methods moving from the seed-and-extend framework used for short reads to a seed-and-chain framework due to the seed abundance in each read. The main novelties are based on alternative seed constructs or chaining formulations. Dozens of tools now exist, whose heuristics have evolved considerably. We provide an overview of the methods used in long-read mappers. Since they are driven by implementation-specific parameters, we develop an original visualization tool to understand the parameter settings ( http://bcazaux.polytech-lille.net/Minimap2/ ).

Although the chemistry of uranium-ligand multiple bonding is burgeoning, analogous complexes involving other actinides such as thorium remain rare and there are not yet any terminal thorium nitrides outside of cryogenic matrix isolation conditions. Here, we report evidence that reduction of a thorium-azide produces a transient Th≡N triple bond, but this activates C-H bonds to produce isolable parent imido derivatives or it can be trapped in an N-heterocycle amine. Computational studies on these thorium-nitrogen multiple bonds consistently evidences a σ > π energy ordering. This suggests pushing-from-below for thorium, where 6p-orbitals principally interact with filled f-orbitals raising the σ-bond energy. Previously this was dismissed for thorium, being the preserve of uranium-nitrides or the uranyl dication. Recognising that pushing-from-below perhaps occurs with thorium as well as uranium, and with imido ligands as well as nitrides, suggests this phenomenon may be more widespread than previously thought.

Emojis are some of the most common ways to convey emotions and sentiments in social messaging applications. In order to help the user choose emojis among a vast range of possibilities, we aim at developing an automatic recommendation system based on user message analysis and real emoji usage, which goes beyond the simple dictionnary lookup that is done in the industry (mainly Android and iOS). For this purpose, we present a novel automatic emoji prediction model trained and tested on real data and based on sentiment-related features. Such a model differ from the ones learnt from tweets and can predict emojis with a 84.48% f1-score and a 95.49% high precision, using MultiLabel-RandomForest algorithm on real private instant message corpus. We want to determine the best discriminative features for this task.

Serious Games (SG) are more and more used for training, as in the crisis management domain, where several hundred stakeholders can be involved, causing various organizational difficulties on field exercises. SGs specific benefits include player immersion and detailed players’ actions tracking during

In this paper we propose an approach that exploits social data associated with a Web resource to measure its a priori relevance. We show how these interaction traces left by the users on the resources, which are in the form of social signals as the number of like and share, can be exploited to quantify social properties such as popularity and reputation. We propose to model these properties as a priori probability that we integrate into language model. We evaluated the effectiveness of our approach on IMDb dataset containing 167438 resources and their social signals collected from several social networks. Our experimental results are statistically significant and show the interest of integrating social properties in a search model to enhance the information retrieval.

In this research, we explore the utilization of the time-delayed feedback method to stabilize unstable steady states and aperiodic orbits within the chaotic fractional-order Rössler Oscillator. Employing the time-delay feedback control algorithm, we identify specific parameter ranges enabling the successful stabilization of unstable equilibria, considering variations in both feedback gain and time delay. Unlike previous research works where Caputo and Riemann-Liouville characterization of fractional derivatives are used, we are using Grünwald-Letnikov (GL) characterization because of its simplicity and ease of implementation and demonstrating the stability using the analytical and numerical analysis and plots of eigenvalues. Additionally, our analysis highlight the effectiveness of a sinusoidally modulated time delay in the control law, significantly expanding the stability region of steady states beyond the capabilities of the traditional time-delayed feedback scheme with a constant delay. Furthermore, the analysis of eigenvalues before and after applying the control strategy offers tangible insights into the system's stability dynamics.

In economics and in perceptual decision-making contextual effects are well documented, where decision weights are adjusted as a function of the distribution of stimuli. Yet, in reinforcement learning literature whether and how contextual information pertaining to decision states is integrated in learning algorithms has received comparably little attention. Here, in an attempt to fill this gap, we investigated reinforcement learning behavior and its computational substrates in a task where we orthogonally manipulated both outcome valence and magnitude, resulting in systematic variations in state-values. Over two experiments, model comparison indicated that subjects’ behavior is best accounted for by an algorithm which includes both reference point-dependence and range-adaptation – two crucial features of state-dependent valuation. In addition, we found state-dependent outcome valuation to progressively emerge over time, to be favored by increasing outcome information and to be correlated with explicit understanding of the task structure. Finally, our data clearly show that, while being locally adaptive (for instance in negative valence and small magnitude contexts), state-dependent valuation comes at the cost of seemingly irrational choices, when options are extrapolated out from their original contexts.

In this paper, we describe a denotational model of Intuitionist Linear Logic which is also a differential category. Formulas are interpreted as Mackey-complete topological vector space and linear proofs are interpreted as bounded linear functions. So as to interpret non-linear proofs of Linear Logic, we use a notion of power series between Mackey-complete spaces, generalizing entire functions in $\mathbb{C}$ . Finally, we get a quantitative model of Intuitionist Differential Linear Logic, with usual syntactic differentiation and where interpretations of proofs decompose as a Taylor expansion.

Fractional-order models provide a powerful framework for capturing memory-dependent and viscoelastic dynamics in mechanical systems, which are often inadequately represented by classical integer-order characterizations. This study addresses the identification of dynamic parameters in both single-degree-of-freedom (1-DOF) and three-degree-of-freedom (3-DOF) Duffing oscillators with fractional damping, modeled using the Grünwald–Letnikov characterization. The 1-DOF system includes a cubic nonlinear restoring force and is excited by a harmonic input to induce steady-state oscillations. For both systems, time domain simulations are conducted to capture long-term responses, followed by Fourier decomposition to extract steady-state displacement, velocity, and acceleration signals. These components are combined with a GL-based fractional derivative approximation to construct structured regressor matrices. System parameters—including mass, stiffness, damping, and fractional-order effects—are then estimated using pseudoinverse techniques. The identified models are validated through a comparison of reconstructed and original trajectories in the phase space, demonstrating high accuracy in capturing the underlying dynamics. The proposed framework provides a consistent and interpretable approach for frequency domain system identification in fractional-order nonlinear systems, with relevance to applications such as mechanical vibration analysis, structural health monitoring, and smart material modeling.

International audience

Abstract In economics and in perceptual decision-making contextual effects are well documented, where decision weights are adjusted as a function of the distribution of stimuli. Yet, in reinforcement learning literature whether and how contextual information pertaining to decision states is integrated in learning algorithms has received comparably little attention. Here, in an attempt to fill this gap, we investigated reinforcement learning behavior and its computational substrates in a task where we orthogonally manipulated both outcome valence and magnitude, resulting in systematic variations in state-values. Over two experiments, model comparison indicated that subjects’ behavior is best accounted for by an algorithm which includes both reference point-dependence and range-adaptation – two crucial features of state-dependent valuation. In addition, we found state-dependent outcome valuation to progressively emerge over time, to be favored by increasing outcome information and to be correlated with explicit understanding of the task structure. Finally, our data clearly show that, while being locally adaptive (for instance in negative valence and small magnitude contexts), state-dependent valuation comes at the cost of seemingly irrational choices, when options are extrapolated out from their original contexts.

Given a rooted, binary phylogenetic network and a rooted, binary phylogenetic tree, can the tree be embedded into the network? This problem, called \textsc{Tree Containment}, arises when validating networks constructed by phylogenetic inference methods.We present the first algorithm for (rooted) \textsc{Tree Containment} using the treewidth $t$ of the input network $N$ as parameter, showing that the problem can be solved in $2^{O(t^2)}\cdot|N|$ time and space.

International audience

International audience

Point patterns are characterized by their density and correlation. While spatial variation of density is well-understood, analysis and synthesis of spatially-varying correlation is an open challenge. No tools are available to intuitively edit such point patterns, primarily due to the lack of a compact representation for spatially varying correlation. We propose a low-dimensional perceptual embedding for point correlations. This embedding can map point patterns to common three-channel raster images, enabling manipulation with off-the-shelf image editing software. To synthesize back point patterns, we propose a novel edge-aware objective that carefully handles sharp variations in density and correlation. The resulting framework allows intuitive and backward-compatible manipulation of point patterns, such as recoloring, relighting to even texture synthesis that have not been available to 2D point pattern design before. Effectiveness of our approach is tested in several user experiments. Code is available at https://github.com/xchhuang/patternshop.

The I4U consortium was established to facilitate a joint entry to NIST speaker recognition evaluations (SRE). The latest edition of such joint submission was in SRE 2018, in which the I4U submission was among the best-performing systems. SRE'18 also marks the 10-year anniversary of I4U consortium into NIST SRE series of evaluation. The primary objective of the current paper is to summarize the results and lessons learned based on the twelve sub-systems and their fusion submitted to SRE'18. It is also our intention to present a shared view on the advancements, progresses, and major paradigm shifts that we have witnessed as an SRE participant in the past decade from SRE'08 to SRE'18. In this regard, we have seen, among others, a paradigm shift from supervector representation to deep speaker embedding, and a switch of research challenge from channel compensation to domain adaptation.

This book presents extended abstracts of the selected contributions to the First International Conference on Advances in Electrical and Computer Engineering (ICAECE'2023), held on 15-16 May 2023 by the Faculty of Science and Technology, Department of Electrical Engineering, University of Echahid Cheikh Larbi Tebessi, Tebessa-Algeria. ICAECE'2023 was delivered in-person and virtually and was open for researchers, engineers, academics, and industrial professionals from around the world interested in new trends and advances in current topics of Electrical and Computer Engineering.

International audience

A fault attack is a malicious manipulation of the hardware (e.g., electromagnetic or laser pulse) that modifies the behavior of the software. Fault attacks typically target sensitive applications such as cryptography services, authentication, boot-loaders or firmware updaters. They can be defended against by adding countermeasures, that is, control flow checks and redundancies, either in the hardware, or in the software running on it. In particular, software countermeasures may be added automatically during compilation. In this paper, we describe a formally verified implementation of this approach in the CompCert verified compiler for the C language. We implemented two existing countermeasures protecting the control flow of the program as program transformations over a middle-end intermediate representation of CompCert, RTL. We proved that these countermeasures are correct, that is, they do not change the observable behavior of the program during an execution without fault injection. We then modeled the effect of a fault on the behavior of the program as an extension of the semantic model of RTL. We used this new model to formally prove the efficacy of the countermeasure: all attacks are either caught, or produce no observable effects. In addition to this formal reasoning, we evaluated the protected program using Lazart, a tool for symbolic fault injection, and measured the effect of optimizations on security and performance.