HAL-CentraleSupelec
Not a member yet
77624 research outputs found
Sort by
Learning-based probabilistic subarray switching for robust low-cost interferometric imaging
No full textInternational audienceComputational cost poses a significant challenge in next-generation interferometric imaging systems. In these systems, the large number of antennas makes it impractical to process all measurements simultaneously due to computational capacity constraints. To reduce the computational burden while preserving image reconstruction quality, we propose a subarray switching strategy that utilizes fewer antennas and different antenna configurations. To take into consideration the influence of the image reconstruction algorithm on the design of the subarray switching pattern and to fully exploit the flexibility of the switching strategy, we propose a probabilistic deep learning-based method for designing antenna switching patterns, named by Probabilistic Antenna Switcher (PAS). In addition to the computational challenge, interferometric systems are also particularly sensitive to the presence of radio frequency interferences (RFI), which heavily affects imaging quality. In order to address this issue, we show that it is possible to combine the proposed PAS with a RFI detection module. Specifically, this module is a neural network that is trained to identify and minimize the impact of RFI-affected antennas in the subarray selection process. This results in a RFI-aware PAS (RaPAS) , which balances computational efficiency, imaging quality, and robustness against RFI
Network Slice Embedding With Flexible Configurations in 5G Networks and Beyond
No full textInternational audienceNetwork slicing enables the creation of multiple virtual networks (i.e., slices) over a shared network infrastructure, each tailored to a specific service. A key challenge lies in network slice embedding, which maps virtual network functions (VNFs) and links onto the physical network. Unlike prior works that assumed fixed configurations, we design a flexible system that allows for selecting the best configuration for each slice based on current physical resource availability during embedding. This leads to a joint optimization of (i) slice configuration selection (SCS) and (ii) slice admission control and embedding (SACE). To solve this, we propose two approaches: an exact method that formulates the joint SCS-SACE problem as an integer linear program (ILP), and a scalable alternative that decouples the problem, solving SCS via reinforcement learning and SACE via either ILP or a heuristic. Simulation results show that allowing flexible configuration selection improves slice acceptance by up to 10%, enabling more efficient slice deployment in resourceconstrained networks
Computation of fuzzy joins over large collections of JSON data using semantic similarity
No full textIn recent years, fuzzy joins have emerged as a critical tool in data integration, data cleaning, and similarity-based retrieval tasks, especially when dealing with noisy or heterogeneous data. Although fuzzy joins have been widely explored for relational data, existing approaches typically focus on textual similarity, matching records based on certain text values and text distances. Apart from relational data, syntactic similarity of XML and JSON documents has also been investigated, where the matching documents are compared over their hierarchical structure. These methods often struggle when confronted with variations in element names, formats, or hierarchical layouts. In contrast, the semantic similarity of semi-structured data, such as JSON documents, remains underexplored, despite its importance in capturing the true meaning conveyed by document values. This paper addresses the problem of performing fuzzy joins between JSON documents by comparing their semantic content rather than their structural form. We propose and describe several methods based on document embeddings, aiming to capture the meaning of a document from its values and contextual relationships. The proposed techniques offer a promising foundation for semantic joining tasks in semi-structured data environments
Token-Efficient Change Detection in LLM APIs
No full textRemote change detection in LLMs is a difficult problem. Existing methods are either too expensive for deployment at scale, or require initial white-box access to model weights or grey-box access to log probabilities. We aim to achieve both low cost and strict black-box operation, observing only output tokens.Our approach hinges on specific inputs we call Border Inputs, for which there exists more than one output top token. From a statistical perspective, optimal change detection depends on the model's Jacobian and the Fisher information of the output distribution. Analyzing these quantities in low-temperature regimes shows that border inputs enable powerful change detection tests.Building on this insight, we propose the Black-Box Border Input Tracking (B3IT) scheme. Extensive in-vivo and in-vitro experiments show that border inputs are easily found for non-reasoning tested endpoints, and achieve performance on par with the best available grey-box approaches. B3IT reduces costs by 30× compared to existing methods, while operating in a strict black-box setting
Solving SIS in any norm via Gaussian sampling
No full textThe short integer solution (SIS) problem is an important problem in lattice-based cryptography. In this paper, we construct a natural and simple algorithm that allows us to solve the problem for any norm in the case where the norm bound ℓ is smaller than half the modulus q. The algorithm consists in using a discrete gaussian sampler on the SIS q-ary lattice to sample lattice vectors, and requires to estimate the probabily that the sampled vector is non-zero and falls into a ball of radius ℓ in the given norm. For the latter, we improve upon previous analysis of random q-ary lattice by obtaining tight bounds on the expected value and variance of the Gaussian mass of the entire lattice and of an ℓp-norm ball, for any p ∈ (0, ∞]. These bounds require new technical results on the discrete Gaussian, but also rely on a conjecture which we have extensively verified. Aside from the conjecture, the remaining part of the algorithm is provably correct. When instantiated with a Markov chain Monte Carlo (MCMC)-based discrete Gaussian sampler, the complexity of the algorithm can be estimated precisely. Although our algorithm does not break Dilithium, it is at least 50 bits faster than the recent algorithm of Ducas, Engelberts and Loyer in Crypto 2025 for all security levels
Towards VR Training Adapted to Affective and Cognitive States: General Method and Evaluation of Mental Workload
No full textInternational audienceiscrete scale (see Fig. 1-B).Abstract— Adaptive virtual reality (VR) applications are used in training and rehabilitation to provide personalized experience, through the adaptive logic that adjust the virtual environment based on the user’s behavior. While users’ behavior is influenced by affective and cognitive states (ACS), adaptive logic typically relies only on users’ performance. First, this paper introduces a general method for the adaptation of VR applications that integrates ACS. We then provide an implementation of this model adapting difficulty with regards to both performance and mental workload. Finally, we present a user study (N=30) to compare our mental workload-based method (experimental) to a state-of-the-art adaptation relying on performance only (control). Results show that our adaptation method led to a decrease of 10.7% in mental workload, an increase of 22.8% in performance, and an overall better experience for most participants. These results were achieved without the participants’ awareness of the adaptive logic of each condition. Taken together, our results promote the integration of ACS in adaptive VR to enhance users’ experience and efficiency, and better fit the function of VR training applications
Modular composition of SPARQL queries for focusing on what to look for rather than how to get it
No full textInternational audienceAdoption of life science knowledge bases by domain experts remains low in spite of the increasing accessibility of these bases, as the Semantic Web framework supports advanced integration and querying. The main bottleneck for leveraging these knowledge bases is that advanced querying combines the inner complexity of life sciences (which requires domain expertise) with the technical complexity of knowledge bases schemas and of SPARQL (which requires engineering skills). We propose a framework based on modules that reconciles both views. A module corresponds to a concept relevant to domain experts and is associated with a SPARQL fragment compliant with the data schema. Modules can be connected to compose new modules corresponding to more complex concepts; the SPARQL fragments of the components are automatically combined to constitute the fragment of the composed module. Our approach thus allows experts to focus on what they are looking for, while our system takes care of how to obtain it
On the Design of an Optimal Multi-Tone Jammer Against the Wiener Interpolation Filter
No full textIn the context of civilian and military communications, anti-jamming techniques are essential to ensure information integrity in the presence of malicious interference. A conventional time-domain approach relies on computing the Wiener interpolation filter to estimate and suppress the jamming waveform from the received samples. It is widely acknowledged that this method is effective for protecting wideband systems against narrowband interference.In this work, this paradigm is questioned through the design of a -tone jamming waveform that is intrinsically difficult to estimate assuming a -tap Wiener interpolation filter. This design relies on an optimization procedure that maximizes the analytical Bayesian mean squared error associated with the jamming waveform estimate. Additionally, an analytical proof is provided showing that a multi-tone jamming waveform composed of tones is sufficient to render the Wiener-filter-based anti-jamming module completely ineffective. The analytical results are validated through Monte Carlo simulations assuming both perfect knowledge and practical estimates of the correlation functions of the received signal
A Berger-Wang formula for impulsive switched systems
No full textThis paper addresses a class of impulsive systems defined by a mix of continuous-time and discrete-time switched linear dynamics. We first analyze a related class of weighted discrete-time switched systems for which we establish a Berger–Wang-type result. An analogous result is then derived for impulsive systems and subsequently used to characterize their exponential stability through a spectral approach, thereby extending existing results in switched-systems theory