Publications

2024

• Decoding Attack Behaviors by Analyzing Patterns in Instruction-Based Attacks using gem5
  
  • Awais Muhammad
  • Mushtaq Maria
  • Naviner Lirida
  • Bruguier Florent
  • Yahya Jawad Haj
  • Benoit Pascal
  , 2024, pp.1-6. The diversity of Instruction Set Architectures (ISAs), each with its unique constraints and optimization strategies, presents significant opportunities and challenges in processor design. Modern processor vendors exploit these ISAs to enhance security, reliability, and performance. Recent security vulnerabilities, notably Spectre and Meltdown, have highlighted the critical need for robust hardware security measures. In this paper, we employ gem5, a state-of-the-art cycle-accurate simulation tool, to simulate the Spectre attack. We developed and modified scripts for both x86 and ARM architectures to ensure compatibility with gem5 version 23.1. Our simulation setup involved running attack scenarios under various configurations to gather comprehensive data on cache misses, cache hits, mispredicted branches, and level 2 cache hits and misses. In the simulation, we analyzed the trace files generated by gem5, utilizing a range of debug flags such as Exec for disassembly (dasm) insights. By detailed analysis of cache and branch prediction using detailed debug data revealed by gem5 traces, we identify some specific attack patterns that are useful for automating the detection of the attacks. Our future work aims to expand this analysis to include additional attack vectors and find more attack patterns, thereby strengthening our attack pattern recognition capabilities. (10.1109/RSP64122.2024.10871078)
  DOI : 10.1109/RSP64122.2024.10871078
• Computational Differential Privacy for Encrypted Databases Supporting Linear Queries
  
  • Alborch Escobar Ferran
  • Canard Sébastien
  • Laguillaumie Fabien
  • Phan Duong Hieu
  Proceedings on Privacy Enhancing Technologies, Privacy Enhancing Technologies Symposium, 2024, 2024 (4), pp.583-604. Differential privacy is a fundamental concept for protecting individual privacy in databases while enabling data analysis. Conceptually, it is assumed that the adversary has no direct access to the database, and therefore, encryption is not necessary. However, with the emergence of cloud computing and the << on-cloud >> storage of vast databases potentially contributed by multiple parties, it is becoming increasingly necessary to consider the possibility of the adversary having (at least partial) access to sensitive databases. A consequence is that, to protect the on-line database, it is now necessary to employ encryption. At PoPETs'19, it was the first time that the notion of differential privacy was considered for encrypted databases, but only for a limited type of query, namely histograms. Subsequently, a new type of query, summation, was considered at CODASPY'22. These works achieve statistical differential privacy, by still assuming that the adversary has no access to the encrypted database. In this paper, we take an essential step further by assuming that the adversary can eventually access the encrypted data, making it impossible to achieve statistical differential privacy because the security of encryption (beyond the one-time pad) relies on computational assumptions. Therefore, the appropriate privacy notion for encrypted databases that we target is computational differential privacy, which was introduced by Beimel et al. at CRYPTO '08. In our work, we focus on the case of functional encryption, which is an extensively studied primitive permitting some authorized computation over encrypted data. Technically, we show that any randomized functional encryption scheme that satisfies simulation-based security and differential privacy of the output can achieve computational differential privacy for multiple queries to one database. Our work also extends the summation query to a much broader range of queries, specifically linear queries, by utilizing inner-product functional encryption. Hence, we provide an instantiation for inner-product functionalities by proving its simulation soundness and present a concrete randomized inner-product functional encryption with computational differential privacy against multiple queries. In terms of efficiency, our protocol is almost as practical as the underlying inner product functional encryption scheme. As evidence, we provide a full benchmark, based on our concrete implementation for databases with up to 1 000 000 entries. Our work can be considered as a step towards achieving privacy-preserving encrypted databases for a wide range of query types and considering the involvement of multiple database owners. (10.56553/popets-2024-0131)
  DOI : 10.56553/popets-2024-0131
• In-Band Sensing and Communication for Optical Access Networks Using Δϕ-OTDR With Simplified Transceivers
  
  • Choudhury Pallab
  • Awwad Élie
  IEEE Sensors Letters, IEEE, 2024, 8 (10), pp.1-4. An in-band integration strategy is proposed by inserting a sensing probe signal over communication data by modulating the same wavelength channel for next-generation optical access network targeting wavelength-division multiplexing (WDM) point-to-point links. The integration is done by exploring the dc-balanced property of a Manchester-coded signal allowing an effective reduction of low-frequency components to accommodate an in-band Golay-coded lower frequency signal that acts as a sensing probe. The system is demonstrated for 10-Gb/s downstream data over a 20-km fiber with a simple direct-detection receiver in a mobile- or enterprise-fronthaul-based WDM link. Differential-phase-sensitive optical time-domain reflectometry is used to locate external perturbations by using the Golay-coded signal for channel estimation and a coherent receiver at the central office. The presented results show that the downstream data can be successfully retrieved from the integrated signal within a pre-forward error correction bit error rate limit of 10−3 maintaining enough input optical power budget at the receiver side. Moreover, the backscattered signal is analyzed for accurate detection of two simultaneous events applied over the fiber maintaining a sensing spatial resolution of 2.1 m and a maximum acoustic bandwidth of 381 Hz with a strain sensitivity down to 15 nϵpp (peak to peak). (10.1109/LSENS.2024.3447091)
  DOI : 10.1109/LSENS.2024.3447091
• Boundary Reconstruction for Wireless Sensor Networks
  
  • Sivadasan Sreeram
  • Govindan Nagarajan
  • Parakkat Amal Dev
  IEEE Access, IEEE, 2024, pp.1-1. Wireless Sensor Networks (WSNs) are critical for various applications ranging from environment monitoring to industrial monitoring. The varying and continuously growing interest in this field demands an understanding of the sensor node distribution to ensure robustness and to improve resource utilization for data processing and decision making. In this paper, we focus on reconstructing the boundaries of a wireless sensor network, which also has a lot of applications in IoT and Robotics. As these sensor node locations can be considered as a set of points in the 2D plane, boundary detection of a WSN can be related to classical shape reconstruction problem in Computational Geometry. In this paper, we extend a simple and generic strategy for hole detection to a geometric solution for boundary/shape reconstruction. Furthermore, we introduce a simple and controllable heuristic algorithm to patch the coverage holes identified by our boundary reconstruction algorithm. Not only does this study improve the reliability of WSNs, but it also provides a useful tool for the extensive domain of computational geometry and shape analysis. Our different experiments show that the proposed reconstruction algorithm outperforms the existing state-of-the-art methods, and hole patching gives a simple and controllable solution for mobile node placement. (10.1109/ACCESS.2024.3483894)
  DOI : 10.1109/ACCESS.2024.3483894
• Refining Wikidata Taxonomy using Large Language Models
  
  • Peng Yiwen
  • Bonald Thomas
  • Alam Mehwish
  , 2024. Due to its collaborative nature, Wikidata is known to have a complex taxonomy, with recurrent issues like the ambiguity between instances and classes, the inaccuracy of some taxonomic paths, the presence of cycles, and the high level of redundancy across classes. Manual efforts to clean up this taxonomy are time-consuming and prone to errors or subjective decisions. We present WiKC, a new version of Wikidata taxonomy cleaned automatically using a combination of Large Language Models (LLMs) and graph mining techniques. Operations on the taxonomy, such as cutting links or merging classes, are performed with the help of zero-shot prompting on an open-source LLM. The quality of the refined taxonomy is evaluated from both intrinsic and extrinsic perspectives, on a task of entity typing for the latter, showing the practical interest of WiKC. (10.1145/3627673.3679156)
  DOI : 10.1145/3627673.3679156
• Weighted Ensemble Models Are Strong Continual Learners
  
  • Marouf Imad Eddine
  • Roy Subhankar
  • Tartaglione Enzo
  • Lathuilière Stéphane
  , 2024, pp.306–324. In this work, we study the problem of continual learning (CL) where the goal is to learn a model on a sequence of tasks, such that the data from the previous tasks becomes unavailable while learning on the current task data. CL is essentially a balancing act between being able to learn on the new task (i.e., plasticity) and maintaining the performance on the previously learned concepts (i.e., stability). Intending to address the stability-plasticity trade-off, we propose to perform weight-ensembling of the model parameters of the previous and current tasks. This weighted-ensembled model, which we call Continual Model Averaging (or CoMA), attains high accuracy on the current task by leveraging plasticity, while not deviating too far from the previous weight configuration, ensuring stability. We also propose an improved variant of CoMA, named Continual Fisher-weighted Model Averaging (or CoFiMA), that selectively weighs each parameter in the weights ensemble by leveraging the Fisher information of the weights of the model. Both variants are conceptually simple, easy to implement, and effective in attaining state-of-the-art performance on several standard CL benchmarks. Code is available at: https://github.com/IemProg/CoFiMA. (10.1007/978-3-031-73209-6_18)
  DOI : 10.1007/978-3-031-73209-6_18
• Memory-Optimized Once-For-All Network
  
  • Girard Maxime
  • Quétu Victor
  • Tardieu Samuel
  • Nguyen Van-Tam
  • Tartaglione Enzo
  , 2025, 15633, pp.252-267. Deploying Deep Neural Networks (DNNs) on different hardware platforms is challenging due to varying resource constraints. Besides handcrafted approaches aiming at making deep models hardware-friendly, Neural Architectures Search is rising as a toolbox to craft more efficient DNNs without sacrificing performance. Among these, the Once-For-All (OFA) approach offers a solution by allowing the sampling of well-performing sub-networks from a single supernet- this leads to evident advantages in terms of computation. However, OFA does not fully utilize the potential memory capacity of the target device, focusing instead on limiting maximum memory usage per layer. This leaves room for an unexploited potential in terms of model generalizability. In this paper, we introduce a Memory-Optimized OFA (MOOFA) supernet, designed to enhance DNN deployment on resource-limited devices by maximizing memory usage (and for instance, features diversity) across different configurations. Tested on ImageNet, our MOOFA supernet demonstrates improvements in memory exploitation and model accuracy compared to the original OFA supernet. Our code is available at https://github.com/MaximeGirard/memory-optimized-once-for-all. (10.1007/978-3-031-91979-4_19)
  DOI : 10.1007/978-3-031-91979-4_19
• Debiasing Surgeon: Fantastic Weights and How to Find Them
  
  • Nahon Rémi
  • de Moura Matos Ivan Luiz
  • Nguyen Van-Tam
  • Tartaglione Enzo
  , 2025, 15143, pp.435-452. Nowadays an ever-growing concerning phenomenon, the emergence of algorithmic biases that can lead to unfair models, emerges. Several debiasing approaches have been proposed in the realm of deep learning, employing more or less sophisticated approaches to discourage these models from massively employing these biases. However, a question emerges: is this extra complexity really necessary? Is a vanilla-trained model already embodying some “unbiased sub-networks” that can be used in isolation and propose a solution without relying on the algorithmic biases? In this work, we show that such a sub-network typically exists, and can be extracted from a vanilla-trained model without requiring additional fine-tuning of the pruned network. We further validate that such specific architecture is incapable of learning a specific bias, suggesting that there are possible architectural countermeasures to the problem of biases in deep neural networks. (10.1007/978-3-031-73013-9_25)
  DOI : 10.1007/978-3-031-73013-9_25
• Privacy-Preserving Adaptive Re-Identification without Image Transfer
  
  • Rami Hamza
  • Giraldo Jhony H.
  • Winckler Nicolas
  • Lathuilière Stéphane
  , 2024, 15110, pp.95-111. Re-Identification systems (Re-ID) are crucial for public safety but face the challenge of having to adapt to environments that differ from their training distribution. Furthermore, rigorous privacy protocols in public places are being enforced as apprehensions regarding individual freedom rise, adding layers of complexity to the deployment of accurate Re-ID systems in new environments. For example, in the European Union, the principles of ``Data Minimization'' and ``Purpose Limitation'' restrict the retention and processing of images to what is strictly necessary. These regulations pose a challenge to the conventional Re-ID training schemes that rely on centralizing data on servers. In this work, we present a novel setting for privacy-preserving Distributed Unsupervised Domain Adaptation for person Re-ID (DUDA-Rid) to address the problem of domain shift without requiring any image transfer outside the camera devices. To address this setting, we introduce Fed-Protoid, a novel solution that adapts person Re-ID models directly within the edge devices. Our proposed solution employs prototypes derived from the source domain to align feature statistics within edge devices. Those source prototypes are distributed across the edge devices to minimize a distributed Maximum Mean Discrepancy (MMD) loss tailored for the DUDA-Rid setting. Our experiments provide compelling evidence that Fed-Protoid outperforms all evaluated methods in terms of both accuracy and communication efficiency, all while maintaining data privacy. (10.1007/978-3-031-72943-0_6)
  DOI : 10.1007/978-3-031-72943-0_6
• Boost Your NeRF: A Model-Agnostic Mixture of Experts Framework for High Quality and Efficient Rendering
  
  • Di Sario Francesco
  • Renzulli Riccardo
  • Tartaglione Enzo
  • Grangetto Marco
  , 2025, 15141, pp.176-192. Since the introduction of NeRFs, considerable attention has been focused on improving their training and inference times, leading to the development of Fast-NeRFs models. Despite demonstrating impressive rendering speed and quality, the rapid convergence of such models poses challenges for further improving reconstruction quality. Common strategies to improve rendering quality involves augmenting model parameters or increasing the number of sampled points. However, these computationally intensive approaches encounter limitations in achieving significant quality enhancements. This study introduces a model-agnostic framework inspired by Sparsely-Gated Mixture of Experts to enhance rendering quality without escalating computational complexity. Our approach enables specialization in rendering different scene components by employing a mixture of experts with varying resolutions. We present a novel gate formulation designed to maximize expert capabilities and propose a resolution-based routing technique to effectively induce sparsity and decompose scenes. Our work significantly improves reconstruction quality while maintaining competitive performance. (10.1007/978-3-031-73010-8_11)
  DOI : 10.1007/978-3-031-73010-8_11
• High Performance Mid-infrared Interband Cascade Lasers on GaSb and Si Substrates
  
  • Kim Hyunah
  • Díaz-Thomas Daniel Andrés
  • Fagot Maëva
  • Spitz Olivier
  • Huang Heming
  • Baranov A. N.
  • Cerutti Laurent
  • Grillot Frédéric
  , 2024, pp.1 - 2. <div><p>A comparison is made between static and dynamic properties of interband cascade lasers grown on GaSb and Si substrates at 3.3 µm. The latter demonstrates promising performance, suggesting future prospects for applications in freespace communications.</p></div> (10.1109/islc57752.2024.10717389)
  DOI : 10.1109/islc57752.2024.10717389
• A Survey of Federative Approaches for Model Management in MBSE
  
  • Amrani Moussa
  • Rakshit Mittal
  • Goulão Miguel
  • Amaral Vasco
  • Guérin Sylvain
  • Martínez Salvador
  • Blouin Dominique
  • Bhobe Anish
  • Hallak Yara
  , 2024. Model-based Systems Engineering (MBSE) advocates the use of models in every stage of development, leading to large number of models that need coordination, collaboration, and discipline management. Model Management (MoM) is a possible approach to manage inter-related collections of models among which Model Federation (MF) provides unique capabilities, like independence of development in individual modelling domains. There is currently a lack of studies about commonalities, variabilities, and gaps in MF approaches. In this paper, we propose a survey and a critical discussion of carefully selected papers about MF. From 59 contributions collected by experts in MoM, we selected, and classified, 23 papers. We extract the main trends we observed, according to our Classification. We then critically review the Classification, and discuss important gaps found in our corpus. The survey results and artefacts are all available online. (10.1145/3652620.3688221)
  DOI : 10.1145/3652620.3688221
• Accuracy enhancement of an optical network digital twin based on open-source field data
  
  • Purkayastha Ambashri
  • Delezoide Camille
  • Lourdiane Mounia
  • Ware Cédric
  • Layec Patricia
  , 2024. <div><p>We propose a two-stage hybrid QoT model for twinning a real transport network and evaluate it on recently published field data. Accounting for partial calibration of key parameters, we improve the SNR prediction accuracy by more than a factor of two. ©2024 The</p></div>
• From Attack Trees to Attack-Defense Trees with Generative AI &amp; Natural Language Processing
  
  • Birchler de Allende Alan
  • Sultan Bastien
  • Apvrille Ludovic
  , 2024, pp.561-569. Attack-defense trees, an extension of attack trees, are extensively used by security engineers to document potential countermeasures for security threats present in a system's design. These trees help integrate initial system models with countermeasures, allowing for early testing of their efficiency and impact in the design cycle. Despite advancements in automating attack tree construction, selecting the initial set of countermeasures for conversion into an attack-defense tree remains largely manual. This paper proposes an approach and a tool that extends the TTool-AI attack tree generation feature by leveraging large language models and natural language processing to create a set of countermeasures and generate attack-defense trees based on an input attack tree. To evaluate our contribution, our approach is tested using attack-defense trees generated from attack trees, each representing possible threats to an associated system specification. In addition, we introduce metrics to assess the semantic correctness and completeness of the generated attack-defense trees. We compared, using our metrics, the attack-defense trees created from our methodology to those created by an engineer and found that attack-defense trees created using AI and secondary mitigation data provided better trees than solely using AI. We also discovered that this approach generated trees that were comparable to the quality of attack-defense trees generated from a security engineer at the associate level. From these results, we believe that our contribution could aid engineers in identifying not only appropriate countermeasures for attack trees but also the optimal number of countermeasures, avoiding the complexity of redundant mitigations. Furthermore, our approach complements standard modeling practices, particularly during the initial design phase, reducing the need for time-consuming re-engineering throughout the system's lifecycle. (10.1145/3652620.3687804)
  DOI : 10.1145/3652620.3687804
• Building Material Permittivity and Conductivity Estimation from 2 to 260 GHz
  
  • Conrat J-M
  • Aliouane Mohamed
  • Cousin Jean-Christophe
  • Begaud Xavier
  , 2024. This paper analyzes the frequency-dependent electromagnetic characteristics of usual flat-surface low-loss building materials. Reflection and transmission losses are measured from 2 to 260 GHz and the related permittivity and conductivity are estimated. Results are in agreement with the ITU-R 2040-3 model that is mainly defined for frequency below 100 GHz. The permittivity can be modelled by a constant value and the conductivity can be modelled by a frequency-dependent function equal to afb. Some differences between the measured and simulated transmission losses have been observed for composite materials that can no longer be considered as homogeneous for frequency above 50 GHz.
• Blind State of Polarisation Monitoring Using Variational AutoEncoders-inspired Adaptive Filter
  
  • tomczyk louis
  • Awwad Élie
  • Prato Diane
  • Ware Cédric
  , 2025. We demonstrate that probabilistically shaped modulations make the Constant Modulus Algorithm unfit for State of Polarisation monitoring in coherent optical communications. Accordingly, we study the potential of Variational AutoEncoders-inspired equalisers for this purpose.
• EPISODIC FINE-TUNING PROTOTYPICAL NETWORKS FOR OPTIMIZATION-BASED FEW-SHOT LEARNING: APPLICATION TO AUDIO CLASSIFICATION
  
  • Zhuang Xuanyu
  • Peeters Geoffroy
  • Richard Gaël
  , 2024. The Prototypical Network (ProtoNet) has emerged as a popular choice in Few-shot Learning (FSL) scenarios due to its remarkable performance and straightforward implementation. Building upon such success, we first propose a simple (yet novel) method to fine-tune a ProtoNet on the (labeled) support set of the test episode of a C-way-K-shot test episode (without using the query set which is only used for evaluation). We then propose an algorithmic framework that combines ProtoNet with optimization-based FSL algorithms (MAML and Meta-Curvature) to work with such a fine-tuning method. Since optimization-based algorithms endow the target learner model with the ability to fast adaption to only a few samples, we utilize ProtoNet as the target model to enhance its fine-tuning performance with the help of a specifically designed episodic fine-tuning strategy. The experimental results confirm that our proposed models, MAML-Proto and MC-Proto, combined with our unique fine-tuning method, outperform regular ProtoNet by a large margin in few-shot audio classification tasks on the ESC-50 and Speech Commands v2 datasets. We note that although we have only applied our model to the audio domain, it is a general method and can be easily extended to other domains.
• WaveTransfer: A Flexible End-to-end Multi-instrument Timbre Transfer with Diffusion
  
  • Baoueb Teysir
  • Bie Xiaoyu
  • Janati Hicham
  • Richard Gael
  , 2024. As diffusion-based deep generative models gain prevalence, researchers are actively investigating their potential applications across various domains, including music synthesis and style alteration. Within this work, we are interested in timbre transfer, a process that involves seamlessly altering the instrumental characteristics of musical pieces while preserving essential musical elements. This paper introduces WaveTransfer, an end-to-end diffusion model designed for timbre transfer. We specifically employ the bilateral denoising diffusion model (BDDM) for noise scheduling search. Our model is capable of conducting timbre transfer between audio mixtures as well as individual instruments. Notably, it exhibits versatility in that it accommodates multiple types of timbre transfer between unique instrument pairs in a single model, eliminating the need for separate model training for each pairing. Furthermore, unlike recent works limited to 16 kHz, WaveTransfer can be trained at various sampling rates, including the industry-standard 44.1 kHz, a feature of particular interest to the music community.
• AI-Driven Consistency of SysML Diagrams
  
  • Sultan Bastien
  • Apvrille Ludovic
  , 2024, pp.149-159. Graphical modeling languages, expected to simplify systems analysis and design, present a challenge in maintaining consistency across their varied views. Traditional rule-based methods for ensuring consistency in languages like UML often fall short in addressing complex semantic dimensions. Moreover, the integration of Large Language Models (LLMs) into Model Driven Engineering (MDE) introduces additional consistency challenges, as LLM's limited output contexts requires the integration of responses. This paper presents a new framework that automates the detection and correction of inconsistencies across different views, leveraging formally defined rules and incorporating OpenAI's GPT, as implemented in TTool. Focusing on the consistency between use case and block diagrams, the framework is evaluated through its application to three case studies, highlighting its potential to significantly enhance consistency management in graphical modeling. (10.1145/3640310.3674079)
  DOI : 10.1145/3640310.3674079
• Mixture of segmentation for heterogeneous functional data
  
  • Brault Vincent
  • Devijver Emilie
  • Laclau Charlotte
  Electronic Journal of Statistics, Shaker Heights, OH : Institute of Mathematical Statistics, 2024, 18 (2), pp.3729-3773. In this paper, we consider functional data with heterogeneity in time and population. We propose a mixture model with segmentation of time to represent this heterogeneity while keeping the functional structure. The maximum likelihood estimator is considered and proved to be identifiable and consistent. In practice, an EM algorithm is used, combined with dynamic programming for the maximization step, to approximate the maximum likelihood estimator. The method is illustrated on a simulated dataset and used on a real dataset of electricity consumption. (10.1214/24-EJS2286)
  DOI : 10.1214/24-EJS2286
• Dynamical behaviour from short to long feedback delay regime in mid-infrared ICL
  
  • Poletti Thomas
  • Kim Hyunah
  • Huang Heming
  • Díaz-Thomas Daniel Andrés
  • Fagot Maëva
  • Baranov A. N.
  • Cerutti Laurent
  • Grillot Frédéric
  , 2024.
• Exploration of Human Repair Initiation in Task-oriented Dialogue : A Linguistic Feature-based Approach
  
  • Ngo Anh
  • Heylen Dirk
  • Rollet Nicolas
  • Pelachaud Catherine
  • Clavel Chloé
  , 2024, pp.603-609. <div><p>In daily conversations, people often encounter problems prompting conversational repair to enhance mutual understanding. By employing an automatic coreference solver, alongside examining repetition, we identify various linguistic features that distinguish turns when the addressee initiates repair from those when they do not. Our findings reveal distinct patterns that characterize the repair sequence and each type of other-repair initiation.</p></div>
• Do Language Models Enjoy Their Own Stories? Prompting Large Language Models for Automatic Story Evaluation
  
  • Chhun Cyril
  • Suchanek Fabian M.
  • Clavel Chloé
  Transactions of the Association for Computational Linguistics, The MIT Press, 2024, 12, pp.1122–1142. Storytelling is an integral part of human experience and plays a crucial role in social interactions. Thus, Automatic Story Evaluation (ASE) and Generation (ASG) could benefit society in multiple ways, but they are challenging tasks which require high-level human abilities such as creativity, reasoning and deep understanding. Meanwhile, Large Language Models (LLM) now achieve state-of-the-art performance on many NLP tasks. In this paper, we study whether LLMs can be used as substitutes for human annotators for ASE. We perform an extensive analysis of the correlations between LLM ratings, other automatic measures, and human annotations, and we explore the influence of prompting on the results and the explainability of LLM behaviour. Most notably, we find that LLMs outperform current automatic measures for system-level evaluation but still struggle at providing satisfactory explanations for their answers. (10.1162/tacl_a_00689)
  DOI : 10.1162/tacl_a_00689
• Invariant Audio Prints for Music Indexing and Alignment
  
  • Mignot Rémi
  • Peeters Geoffroy
  , 2024, pp.1-7. This work deals with music indexing and alignment using audio codes designed to be representative of the music content and robust to sound modifications. First, based on properties of the Fourier Transform and of the logarithm, high-dimensional audio descriptors are designed. Then, a dimension reduction is learned with criteria based on sound discrimination and invariance to transformations. Finally, a binarization is computed to derive codes (integers). This last process allows a fast searching for large catalogs with a hash table, and a Hamming distance on codes makes possible the time alignment using an adapted "Dynamic Time Warping". The contributions of this paper are tested for two different tasks. The goal of the first task is to identify the segments of music medleys with the audio indexing process, and to accurately find the corresponding original time positions. The goal of the second task is to measure the accuracy of the time-alignment with synthesized MIDI files, where the tempo continuously varies, and with modified pitches and instruments. Additionally, the audio indexing is also tested for these data, in order to exhibit some properties of the used audio prints. (10.1109/CBMI62980.2024.10859214)
  DOI : 10.1109/CBMI62980.2024.10859214
• Riemannian optimization of photonic quantum circuits in phase and Fock space
  
  • Yao Yuan
  • Miatto Filippo
  • Quesada Nicolás
  SciPost Physics, SciPost Foundation, 2024. <div><p>We propose a framework to design and optimize generic photonic quantum circuits composed of Gaussian objects (pure and mixed Gaussian states, Gaussian unitaries, Gaussian channels, Gaussian measurements) as well as non-Gaussian effects such as photonnumber-resolving measurements. In this framework, we parametrize a phase space representation of Gaussian objects using elements of the symplectic group (or the unitary or orthogonal group in special cases), and then we transform it into the Fock representation using a single linear recurrence relation that computes the Fock amplitudes of any Gaussian object recursively. We also compute the gradient of the Fock amplitudes with respect to phase space parameters by differentiating through the recurrence relation. We can then use Riemannian optimization on the symplectic group to optimize M-mode Gaussian objects, avoiding the need to commit to particular realizations in terms of fundamental gates. This allows us to "mod out" all the different gate-level implementations of the same circuit, which now can be chosen after the optimization has completed. This can be especially useful when looking to answer general questions, such as bounding the value of a property over a class of states or transformations, or when one would like to worry about hardware constraints separately from the circuit optimization step. Finally, we make our framework extendable to non-Gaussian objects that can be written as linear combinations of Gaussian ones, by explicitly computing the change in global phase when states undergo Gaussian transformations. We implemented all of these methods in the freely available open-source library MrMustard [1], which we use in three examples to optimize the 216-mode interferometer in Borealis, and 2-and 3-modes circuits (with Fock measurements) to produce cat states and cubic phase states.</p></div> (10.21468/SciPostPhys)
  DOI : 10.21468/SciPostPhys