Publications

2023

• Glass-to-glass delay reduction: encoding rate reduction vs. video frame extrapolation
  
  • Kanj Hind
  • Trioux Anthony
  • Cagnazzo Marco
  • Coudoux François-Xavier
  • Corlay Patrick
  • Kieffer M.
  , 2023, pp.1-6. Applications such as teleoperated driving, remote robot control, and telepresence rely on video services to ensure real-time interaction with a satisfying quality of experience. Reducing the Glass-to-Glass (G2G) delay, i.e., the time delay between the acquisition of a video frame and its display on a remote terminal is critical for these applications. Deep learning-based video frame extrapolation before video encoding has been recently considered as an interesting solution to reduce G2G delay, however, the latency introduced by extrapolation has not been taken into account. In this paper, considering the main sources of latency, including extrapolation delay, we examine the benefits and limitations of frame extrapolation at encoder in reducing the G2G delay in a point-to-point video transmission system. To this end, we compare the latency-quality trade-off for two latency compensation methods: encoding rate reduction and video frame extrapolation. Our aim is to determine the G2G delay reduction that may be achieved at the price of a given quality reduction. Our experiments show that extrapolation methods can provide a null perceived G2G delay with an acceptable loss in quality, particularly for applications with video contents with limited temporal information. Such delay reduction is unreachable via encoding rate reduction. (10.1109/MMSP59012.2023.10337718)
  DOI : 10.1109/MMSP59012.2023.10337718
• Surveying the Social Comfort of Body, Device, and Environment-Based Augmented Reality Interactions in Confined Passenger Spaces Using Mixed Reality Composite Videos
  
  • Medeiros Daniel
  • Dubus Romane
  • Williamson Julie
  • Wilson Graham
  • Pöhlmann Katharina
  • Mcgill Mark
  Proceedings of the ACM on Interactive, Mobile, Wearable and Ubiquitous Technologies, ACM, 2023, 7 (3), pp.1 - 25. Augmented Reality (AR) headsets could significantly improve the passenger experience, freeing users from the restrictions of physical smartphones, tablets and seatback displays. However, the confined space of public transport and the varying proximity to other passengers may restrict what interaction techniques are deemed socially acceptable for AR users - particularly considering current reliance on mid-air interactions in consumer headsets. We contribute and utilize a novel approach to social acceptability video surveys, employing mixed reality composited videos to present a real user performing interactions across different virtual transport environments. This approach allows for controlled evaluation of perceived social acceptability whilst freeing researchers to present interactions in any simulated context. Our resulting survey (N=131) explores the social comfort of body, device, and environment-based interactions across seven transit seating arrangements. We reflect on the advantages of discreet inputs over mid-air and the unique challenges of face-to-face seating for passenger AR. (10.1145/3610923)
  DOI : 10.1145/3610923
• The challenging neural decoding with general-purpose networks and its improvement via probabilistic embeddings
  
  • Wang Xiaolin
  • Boutros Joseph J.
  • Rioul Olivier
  , 2023. As deep learning is becoming more and more popular in a variety of fields, it is natural to ask whether it can be used for decoding error-correcting codes. In this work, we show that the answer is yes, but with a caveat: Naive application of general-purpose neural networks is not well suited for bitwise decoding. We thoroughly identify the challenges that prevent general-purpose neural networks from decoding successfully, including the curse of dimensionality and the requirement of extremely high accuracy. We then propose a proba- bilistic embedding method in the preprocessing stage that facilitate the learning. We also show that this new method allows general-purpose neural networks to decode with a perfor- mance that is close to the theoretical optimality while saving time compared to traditional decoding methods such as maximum-likelihood decoding or the BCJR algorithm.
• Rogue waves and extreme events in mid infrared quantum cascade lasers under external optical feedback
  
  • Grillot Frédéric
  • Spitz Olivier
  , 2023. Quantum cascade lasers (QCLs) are optical sources exploiting radiative intersubband transitions within the conduction band of semiconductor heterostructures, which further guide the electron flow by tunneling and scattering to establish electronic inversion for a pair of quantum levels at a given electric field [1]. In order to increase the total gain, a module of several layers including the laser levels is repeated several times (Fig. 1 (left)), so that the electrons traverse the total structure like water in a cascade. Emission range of QCLs typically extends from the midinfrared to the terahertz region hence making them candidates of choice for a wide range of applications such as free-space laser communications [2] or optical radars [3]. This work aims at exploring rogue waves and optical dragon-kings arising in QCLs under optical feedback. While it is obvious that disruptive events will affect a transmission link, the detection and suppression of rogue waves is important for improving free-space data transmission. Ultimately, the control of these extreme events to a level such that a QCL could be used as a rogue waves generator could even be utilized to disrupt a free-space transmission link.
• Quantum Photonics with Interband Cascade Lasers
  
  • Zhao Shiyuan
  • Grillot Frédéric
  , 2023. Quantum photonics with interband cascade lasers (ICLs) is an emerging research field that investigates the use of ICLs for quantum communication, quantum cryptography, and other applications that require the generation and manipulation of quantum states. In this work, we explore the possibility of generating amplitudesqueezed light with high quantum efficiency ICLs through theoretical investigations. Based on two different modelling approaches, we demonstrate that this mid-infrared semiconductor device can operate with significant amplitude squeezing over a large bandwidth of several GHz when driven by low-noise constant current sources. Our findings could accelerate the development of original quantum hardware in the mid-infrared range, which is currently not available but could have numerous applications, including laser-based free-space secure communication systems.
• Optimization of security risk for learning on heterogeneous quality data
  
  • Chaitou Hassan
  , 2023. Intrusion Detection Systems (IDSs) serve as critical components in network security infrastructure. In order to cope with the scalability issues of IDSs using handcrafted detection rules, machine learning is used to design IDSs trained on datasets. Yet, they are increasingly challenged by meta-attacks, called adversarial evasion attacks, that alter existing attacks to improve their evasion capabilities. These approaches, for instance, employ Generative Adversarial Networks (GANs) to automate the alteration process. Several strategies have been proposed to enhance the robustness of IDSs against such attacks, with significant success in strategies based on adversarial training. However, IDSs evasion remains relevant as many contributions also show that adversarial evasion attacks are still efficient despite using adversarial training on IDSs. In this thesis, we investigate this situation and present contributions that improve the understanding of one of its root causes and guidelines to mitigate it. The first step is to better understand the possible sources of variability in IDS or evasion attack performances. Three potential sources are considered: methodological assessment issues, the inherent race to spend more computational resources in attack or defense, or issues in training and dataset acquisition when training IDSs. The first contribution consists of guidelines to conduct robust IDSs assessments beyond the simple recommendation for empirical analysis. These guidelines cover both single experiment design but also sensitivity analysis campaigns. The consequence of applying such guidelines is to obtain more stable results when changing training resource related parameters. Removing artifacts due to inadequate assessment procedures leads us to investigate why some selected parts of the considered dataset tend to be almost not affected by adversarial attacks. The second contribution is the formalization of adversarial neighborhoods: an alternative way to characterize adversarial samples. This formalization allows us to adapt and evaluate data quality criteria used for non-adversarial samples, such as the absence of contradictory samples, and apply similar criteria to adversarial sample datasets. From this concept, four threat situations have been identified with clear qualitative impacts either on the training of a robust IDS or the attacker’s ability to find more successful evasion attacks. Finally, we propose countermeasures to the identified threats and then perform an empirical quantitative assessment of both threats and countermeasures. The findings of these experiments highlight the need to identify and mitigate threats associated with a non-empty extended contradictory set. Indeed, this crucial vulnerability should be identified and addressed prior to IDS training.
• Challenges in Generating True Random Numbers Considering the Variety of Corners, Aging, and Intentional Attacks
  
  • Bahrami Javad
  • Danger Jean-Luc
  • Ebrahimabadi Mohammad
  • Guilley Sylvain
  • Karimi Naghmeh
  , 2023, pp.10-15. True Random Number Generators (TRNGs) are sensitive Intellectual Property (IP) blocks involved in the creation of cryptographic keys, initialization vectors, nonces, etc. They must behave properly within a large environmental spectrum, including multiple corners, in case of aging-induced change of device characteristics over time, and also under intentional attacks aiming at lowering the TRNGs entropy. In this paper, we review normative and technical landscapes in this respect, and propose a pre-silicon verification methodology to assess the resilience of TRNGs. In particular, we qualify the unitary free running oscillator (FRO) entropy source analytically, and then extend the study to a full FRO-based IP module. Our results encompass analytical characterizations in terms of jitter measurements and certification-based characterizations in terms of tests. (10.1109/ICICDT59917.2023.10332319)
  DOI : 10.1109/ICICDT59917.2023.10332319
• A Data-Driven Approach for Modeling Unknown Multi-Scale Systems
  
  • Pol Marius
  • Diaconescu Ada
  , 2023, ACSOS-C 2023, pp.35-40. Complex adaptive systems often organize via multiple abstraction levels, or ‘scales’, interconnected by feedback loops. This enables adaptation and survival in changing environments, while managing complexity with limited resources. For an external observer unaware of such multi-scale structure, modeling an unknown system may be a complicated endeavor. This position paper proposes a data-driven approach for addressing this issue. It generates multi-scale models from incomplete monitoring data, capitalizing on the behavioral regularities that stem from its feedback loops. It also defines the appropriate language elements for expressing these multi-scale models. We validate our approach on data obtained from a theoretical multi-scale system: a holonic cellular automata (HCA) simulator. Results show that the proposed approach can identify the HCA's three abstraction levels and main modeling concepts. This is an encouraging first step towards establishing automatic methods for multi-scale model discovery from partial observations. (10.1109/ACSOS-C58168.2023.00033)
  DOI : 10.1109/ACSOS-C58168.2023.00033
• Equalization in dispersion-managed systems using learned digital back-propagation
  
  • Abu-Romoh Mohannad
  • Costa Nelson
  • Jaouën Yves
  • Napoli Antonio
  • Pedro João
  • Spinnler Bernhard
  • Yousefi Mansoor
  Optics Continuum, Optica Publishing Group, 2023, 2 (10), pp.2088. In this paper, we investigate the use of the learned digital back-propagation (LDBP) for equalizing dual-polarization fiber-optic transmission in dispersion-managed (DM) links. LDBP is a deep neural network that optimizes the parameters of DBP using the stochastic gradient descent. We evaluate DBP and LDBP in a simulated WDM dual-polarization fiber transmission system operating at 32 Gbaud/s per channel, with a dispersion map designed for a 28 × 72 km link with 15% residual dispersion. Our results show that in single-channel transmission, LDBP achieves an effective signal-to-noise ratio improvement of 6.3 dB and 2.5 dB using DP-16-QAM format, respectively, over linear equalization and DBP. In WDM transmission, the corresponding Q -factor gains are 1.1 dB and 0.4 dB, respectively. Additionally, we conduct a complexity analysis, which reveals that a frequency-domain implementation of LDBP and DBP is more favorable in terms of complexity than the time-domain implementation. These findings demonstrate the effectiveness of LDBP in mitigating the nonlinear effects in DM fiber-optic transmission systems. (10.1364/OPTCON.497117)
  DOI : 10.1364/OPTCON.497117
• Hierarchy Beyond Top-Down Control: The Architecture of Self-Organised Social Systems
  
  • Di Felice Louisa Jane
  • Diaconescu Ada
  , 2023, 2023, pp.80-85. Diverse social system architectures are needed to address the inter-related social and environmental challenges that societies face across scales. The iron law of oligarchy states that all societies eventually become hierarchical, with top-down control, as they grow in size and complexity. We contribute to existing research arguing that this is not an inevitability. To do this, we analyse social systems that organise non-hierarchically (Self-Organised Social Systems (SOSS)). SOSS have a multi-scale architecture with respect to information abstraction, which is not tied to a hierarchical control structure, and allows for scalability. Through primary and secondary data collection, we analyse the structure and function of SOSS in the past and present, including neighbourhood collectives, activist and co-living groups (primary data, via interviews); and anarchist collectives in the Spanish Revolution, the Zapatistas, and the Occupy and 15M movements, among others (secondary data, from literature). We identify federations and networks as two initial types of SOSS that can scale in size and complexity. Key features include: (i) maintaining power at the lowest level; (ii) the quasi-autonomy of lower-level groups; (iii) sharing of skills, information and knowledge; (iv) adaptability across scales; (v) stability and uncertainty reduction; (vi) resilience. Future work will focus on questions of timing, scalar stress, and adaptability, to better understand how and why SOSS architectures succeed, fail, and can be implemented. (10.1109/ACSOS-C58168.2023.00042)
  DOI : 10.1109/ACSOS-C58168.2023.00042
• Speech Self-Supervised Representations Benchmarking: a Case for Larger Probing Heads
  
  • Zaiem Salah
  • Kemiche Youcef
  • Parcollet Titouan
  • Essid Slim
  • Ravanelli Mirco
  , 2023. Self-supervised learning (SSL) leverages large datasets of unlabeled speech to reach impressive performance with reduced amounts of annotated data. The high number of proposed approaches fostered the emergence of comprehensive benchmarks that evaluate their performance on a set of downstream tasks exploring various aspects of the speech signal. However, while the number of considered tasks has been growing, most proposals rely upon a single downstream architecture that maps the frozen SSL representations to the task labels. This study examines how benchmarking results are affected by changes in the probing head architecture. Interestingly, we found that altering the downstream architecture structure leads to significant fluctuations in the performance ranking of the evaluated models. Against common practices in speech SSL benchmarking, we evaluate larger-capacity probing heads, showing their impact on performance, inference costs, generalization and multi-level feature exploitation.
• Training dynamic models using early exits for automatic speech recognition on resource-constrained devices
  
  • Wright George August
  • Cappellazzo Umberto
  • Zaiem Salah
  • Raj Desh
  • Yang Lucas Ondel
  • Falavigna Daniele
  • Brutti Alessio
  , 2023. The possibility of dynamically modifying the computational load of neural models at inference time is crucial for on-device processing, where computational power is limited and time-varying. Established approaches for neural model compression exist, but they provide architecturally static models. In this paper, we investigate the use of early-exit architectures, that rely on intermediate exit branches, applied to large-vocabulary speech recognition. This allows for the development of dynamic models that adjust their computational cost to the available resources and recognition performance. Unlike previous works, besides using pre-trained backbones we also train the model from scratch with an early-exit architecture. Experiments on public datasets show that early-exit architectures from scratch not only preserve performance levels when using fewer encoder layers, but also improve task accuracy as compared to using single-exit models or using pre-trained models. Additionally, we investigate an exit selection strategy based on posterior probabilities as an alternative to frame-based entropy.
• Overview of Touché 2023: Argument and Causal Retrieval
  
  • Bondarenko Alexander
  • Fröbe Maik
  • Kiesel Johannes
  • Schlatt Ferdinand
  • Barriere Valentin
  • Ravenet Brian
  • Hemamou Léo
  • Luck Simon
  • Reimer Jan Heinrich
  • Stein Benno
  • Potthast Martin
  • Hagen Matthias
  , 2023, 14163, pp.507-530. This paper is a condensed overview of Touché: the fourth edition of the lab on argument and causal retrieval that was held at CLEF 2023. With the goal to create a collaborative platform for research on computational argumentation and causality, we organized four shared tasks: (a) argument retrieval for controversial topics, where participants retrieve web documents that contain high-quality argumentation and detect the argument stance, (b) causal retrieval, where participants retrieve documents that contain causal statements from a generic web crawl and detect the causal stance, (c) image retrieval for arguments, where participants retrieve from a focused web crawl images showing support or opposition to some stance, and (d) multilingual multi-target stance classification, where participants detect the stance of comments on proposals from an online multilingual participatory democracy platform. (10.1007/978-3-031-42448-9_31)
  DOI : 10.1007/978-3-031-42448-9_31
• ORSUM 2023 - 6th Workshop on Online Recommender Systems and User Modeling
  
  • Vinagre João
  • Al-Ghossein Marie
  • Peska Ladislav
  • Jorge Alípio Mário
  • Bifet Albert
  , 2023, pp.1272--1273. Modern online platforms for user modeling and recommendation require complex data infrastructures to collect and process data. Some of this data has to be kept to later be used in batches to train personalization models. However, since user activity data can be generated at very fast rates it is also useful to have algorithms able to process data streams online, in real time. Given the continuous and potentially fast change of content, context and user preferences or intents, stream-based models, and their synchronization with batch models can be extremely challenging. Therefore, it is important to investigate methods able to transparently and continuously adapt to the inherent dynamics of user interactions, preferably over long periods of time. Models able to continuously learn from such flows of data are gaining attention in the recommender systems community, and are being increasingly deployed in online platforms. However, many challenges associated with learning from streams need further investigation. The objective of this workshop is to foster contributions and bring together a growing community of researchers and practitioners interested in online, adaptive approaches to user modeling, recommendation and personalization, and their implications regarding multiple dimensions, such as reproducibility, privacy, fairness, diversity, transparency, auditability, and compliance with recently adopted or upcoming legal frameworks worldwide. (10.1145/3604915.3608763)
  DOI : 10.1145/3604915.3608763
• FG\(²\)AN: Fairness-Aware Graph Generative Adversarial Networks
  
  • Wang Zichong
  • Wallace Charles
  • Bifet Albert
  • Yao Xin
  • Zhang Wenbin
  , 2023, 14170, pp.259--275. Graph generation models have gained increasing popularity and success across various domains. However, most research in this area has concentrated on enhancing performance, with the issue of fairness remaining largely unexplored. Existing graph generation models prioritize minimizing graph reconstruction’s expected loss, which can result in representational disparities in the generated graphs that unfairly impact marginalized groups. This paper addresses this socially sensitive issue by conducting the first comprehensive investigation of fair graph generation models by identifying the root causes of representational disparities, and proposing a novel framework that ensures consistent and equitable representation across all groups. Additionally, a suite of fairness metrics has been developed to evaluate bias in graph generation models, standardizing fair graph generation research. Through extensive experiments on five real-world datasets, the proposed framework is demonstrated to outperform existing benchmarks in terms of graph fairness while maintaining competitive prediction performance. (10.1007/978-3-031-43415-0_16)
  DOI : 10.1007/978-3-031-43415-0_16
• Time-Domain Traveling-Wave Model of Distributed-feedback Quantum Cascade Laser
  
  • Zaminga Sara
  • Columbo Lorenzo
  • Silvestri Carlo
  • Gioannini Mariangela
  • Grillot Frédéric
  , 2023, pp.9-10. In this paper, a time-domain (TD) traveling-wave (TW) model based on the Effective Semiconductor Maxwell-Bloch Equations (ESMBEs) and a coupled-mode theory is proposed for the description of the dynamics of a mid-Infrared (mid-IR) Quantum Cascade Laser (QCL) in the Distributed-Feedback (DFB) configuration. The influence of physical and geometrical properties on the QCL’s dynamics is investigated. Numerical simulations find good agreement with the experimental results obtained with a DFB QCL operating at 9.34 μm. (10.1109/NUSOD59562.2023.10273563)
  DOI : 10.1109/NUSOD59562.2023.10273563
• NeMig - A Bilingual News Collection and Knowledge Graph about Migration
  
  • Iana Andreea
  • Alam Mehwish
  • Grote Alexander
  • Luwig Katharina
  • Müller Philipp
  • Weinhardt Christof
  • Paulheim Heiko
  , 2023. News recommendation plays a critical role in shaping the public's worldviews through the way in which it filters and disseminates information about different topics. Given the crucial impact that media plays in opinion formation, especially for sensitive topics, understanding the effects of personalized recommendation beyond accuracy has become essential in today's digital society. In this work, we present NeMig, a bilingual news collection on the topic of migration, and corresponding rich user data. In comparison to existing news recommendation datasets, which comprise a large variety of monolingual news, NeMig covers articles on a single controversial topic, published in both Germany and the US. We annotate the sentiment polarization of the articles and the political leanings of the media outlets, in addition to extracting subtopics and named entities disambiguated through Wikidata. These features can be used to analyze the effects of algorithmic news curation beyond accuracy-based performance, such as recommender biases and the creation of filter bubbles. We construct domain-specific knowledge graphs from the news text and metadata, thus encoding knowledge-level connections between articles. Importantly, while existing datasets include only click behavior, we collect user socio-demographic and political information in addition to explicit click feedback. We demonstrate the utility of NeMig through experiments on the tasks of news recommenders benchmarking, analysis of biases in recommenders, and news trends analysis. NeMig aims to provide a useful resource for the news recommendation community and to foster interdisciplinary research into the multidimensional effects of algorithmic news curation. (10.48550/arXiv.2309.00550)
  DOI : 10.48550/arXiv.2309.00550
• Broadband amplitude squeezing in electrically driven quantum dot lasers
  
  • Zhao Shiyuan
  • Ding Shihao
  • Huang Heming
  • Zaquine Isabelle
  • Fabre Nicolas
  • Grillot Frédéric
  • Belabas Nadia
  , 2023. The generation of broadband squeezed states of light lies at the heart of high-speed continuousvariable quantum information. Traditionally, optical nonlinear interactions have been employed to produce quadrature-squeezed states. However, the harnessing of electrically pumped semiconductor lasers offers distinctive paradigms to achieve enhanced squeezing performance. We present evidence that quantum dot lasers enable the realization of broadband amplitude-squeezed states at room temperature across a wide frequency range, spanning from 3 GHz to 12 GHz. Our findings are corroborated by a comprehensive stochastic simulation in agreement with the experimental data.
• AMULET: a Mutation Language Enabling Automatic Enrichment of SysML Models
  
  • Sultan Bastien
  • Frénot Léon
  • Apvrille Ludovic
  • Jaillon Philippe
  • Coudert Sophie
  ACM Transactions on Embedded Computing Systems (TECS), ACM, 2023, pp.1-28. SysML models are widely used for designing and analyzing complex systems. Model-based design methods often require successive modifications of the models, whether for incrementally refining the design (e.g. in agile development methods) or for testing different design options. Such modifications, or mutations, are also used in mutation-based testing approaches. However, the definition of mutation operators can be a complex issue and applying them to models is sometimes performed by hand: this is time consuming and error prone. The paper addresses this issue thanks to the introduction of AMULET, the first mutation language for SysML. AMULET encompasses the modifications targeting SysML block and state-machine diagrams, and is supported by a compiler the paper presents. This compiler is integrated in TTool, an open-source SysML toolkit, enabling the full support of design methods including model design, mutation and verification tasks in a unique toolkit. The paper also introduces two case-studies providing concrete examples of AMULET use for modeling vulnerabilities and cyber attacks, and highlighting the benefits of AMULET for SysML mutations. (10.1145/3624583)
  DOI : 10.1145/3624583
• Deep Reinforcement Learning for Resource Constrained Multiclass Scheduling in Wireless Networks
  
  • Avranas Apostolos
  • Ciblat Philippe
  • Kountouris Marios
  IEEE Transactions on Machine Learning in Communications and Networking, Institute of Electrical and Electronics Engineers, 2023, 1, pp.225-241. The problem of multiclass scheduling in a dynamic wireless setting is considered here, where the available limited bandwidth resources are allocated to handle random service demand arrivals, belonging to different classes in terms of payload data request, delay tolerance, and importance/priority. In addition to heterogeneous traffic, another major challenge stems from random service rates due to time-varying wireless communication channels. Existing scheduling and resource allocation approaches, ranging from simple greedy heuristics and constrained optimization to combinatorics, are tailored to specific network or application configuration and are usually suboptimal. On this account, we resort to deep reinforcement learning (DRL) and propose a distributional Deep Deterministic Policy Gradient (DDPG) algorithm combined with Deep Sets to tackle the aforementioned problem. Furthermore, we present a novel way to use a Dueling Network, which leads to further performance improvement. Our proposed algorithm is tested on both synthetic and real data, showing consistent gains against baseline methods from combinatorics and optimization, and state-of-the-art scheduling metrics. Our method can, for instance, achieve with 13% less power and bandwidth resources the same user satisfaction rate as a myopic algorithm using knapsack optimization (10.1109/TMLCN.2023.3314705)
  DOI : 10.1109/TMLCN.2023.3314705
• When to generate hedges in peer-tutoring interactions
  
  • Abulimiti Alafate
  • Clavel Chloé
  • Cassell Justine
  , 2023. This paper explores the application of machine learning techniques to predict where hedging occurs in peer-tutoring interactions. The study uses a naturalistic face-to-face dataset annotated for natural language turns, conversational strategies, tutoring strategies, and nonverbal behaviours. These elements are processed into a vector representation of the previous turns, which serves as input to several machine learning models. Results show that embedding layers, that capture the semantic information of the previous turns, significantly improves the model's performance. Additionally, the study provides insights into the importance of various features, such as interpersonal rapport and nonverbal behaviours, in predicting hedges by using Shapley values for feature explanation. We discover that the eye gaze of both the tutor and the tutee has a significant impact on hedge prediction. We further validate this observation through a follow-up ablation study.
• Quantization of Neural Network Equalizers in Optical Fiber Transmission Experiments
  
  • Darweesh Jamal
  , 2023. The advent of the coherent detection paved the away for the compensation of the fibertransmission effects in the electrical domain using the digital signal processing (DSP).While the chromatic and polarization mode dispersion can be efficiently compensated withDSP, the compensation of the nonlinear distortions remains challenging.In this work, we consider neural networks (NNs) for nonlinearity mitigation in dualpolarization optical fiber transmission. Compared to the digital back-propagation (DBP),NNs do not require the fiber link parameters, and may mitigate the impairments withlower complexity.We propose two low-complexity NN equalizers: a convolutional-dense and an LSTM-dense model, placed at the end of the linear DSP to compensate the nonlinearities. Theseequalizers are evaluated in the context of three dual-polarization transmission experiments:a 9x50km true-wave classic fiber link, a 9x110km standard single-mode fiber link, and a17x70km LEAF fiber link. It is shown that the proposed NNs and DBP achieve about thesame Q-factors, both outperforming the linear DSP.We use quantization in order to reduce the computational complexity, storage sizeand energy consumption of the NN equalizers. We compare a number of post-trainingquantization (PTQ) and training-aware quantization (TAQ) algorithms for casting theweights and activations of the NN in few bits. For quantization above 5 bits, we showthat TAQ with straight-through estimation (STE) outperforms PTQ, since it mitigatesthe quantization noise during the training to some extent. For a Q-factor drop of less than0.5 dB compared to the unquantized NN, the storage and computational complexity of theNN can be typically reduced by over 90%. However, there is a bit width cut-off value ofaround 5 bits below which TAQ fails to outperform the linear DSP. This is because, theapproximation of the derivative of the quantizer in the STE is not sufficiently accurate atlow bit widths. Further, the proposed low-complexity models are not overparameterized,so that the quantization noise can be mitigated during the training at low bit widths. Itis shown that the quantization of the activations has a greater impact on the performancecompared to the quantization of the weights.Finally, we study extreme quantization of the NN equalizers below 5 bits. For thiscase, we propose three novel algorithms: successive PTQ (SPTQ), alpha-blending (AB)and successive AB (SAB) which is a hybrid algorithm that combines the SPTQ with AB.These algorithms are iterative, and incorporate ideas from PTQ and TAQ. We demonstrateiiithat the weights of the NN can be quantized up to one bit, if the activations are notquantized. Further, it is shown that both weights and activations can be quantized at 2-3bits, while still notably outperforming the linear equalization. Furthermore, we quantifythe impact of the quantization noise arising separately from the weights and activationson the Q-factor performance of the model. The results demonstrate for the first time thatlow-complexity binary NNs can mitigate nonlinearities in optical fiber communication.This PhD thesis is in the frame of a European Union's Horizon 2020 MSCA-ITN-EID REAL-NET project, grant agreement no. 813144, in collaboration with Infinera inGermany and Portugal
• Local sampling of the SU(1,1) Wigner function
  
  • Fabre Nicolas
  • Klimov Andrei
  • Leuchs Gerd
  • Sanchez Soto Luis
  , 2023, 287, pp.06016. The Wigner phase-space formulation for systems possessing SU(1,1) symmetry has been defined by Seyfarth et al. [Quantum 4 , 317 (2020)] tackling the difficulty in defining a suitable operational definition of the Wigner function. To further investigate this formulation, we propose a non-linear optical setup that incorporates photon-number-resolving detectors, which would enable a direct and comprehensive point-by-point sampling of the SU(1,1) Wigner function. We discuss the visualization of various two-mode quantum states and the effect of the losses in such a detection scheme. (10.1051/epjconf/202328706016)
  DOI : 10.1051/epjconf/202328706016
• A hitchhiker's guide to white-box neural network watermarking robustness
  
  • de Sousa Trias Carl
  • Mitrea Mihai P
  • Tartaglione Enzo
  • Fiandrotti Attilio
  • Cagnazzo Marco
  • Chaudhuri Sumanta
  , 2023. The present study deals with white-box Neural Network (NN) watermarking and focuses on the robustness property. The first contribution consists of formalizing neuron permutation as a geometric attack, thus demonstrating the very existence of this class of attacks for NN watermarking. The second contribution consists in devising and demonstrating the effectiveness of the corresponding counterattack. As a side result, the possibility of extending NN white-box watermarking scope beyond image classification is brought to light. The experimental study considers three state-of-the-art methods, four NN models, three tasks (image classification, segmentation, and video coding), and five types of attacks. We underline that none of the existing methods is robust against the geometric attack, and using the counterattack advanced in this paper effectively ensures the robustness.
• Software Defined Radio platform to evaluate processing latency of 5G NR MIMO functions
  
  • Caloyannis Karen
  • Vergne Anaïs
  • Martins Philippe
  , 2023. This paper presents a Software Defined Radio (SDR) implementation in C++ of a Multiple Input Multiple Output (MIMO) transceiver platform using the open-source 5G physical layer free5GRAN. The platform evaluates the processing latency of physical layer functions at the receiver on a x86 processor, required for 2 and 4 layer MIMO transmissions. The goal is to verify if the functions can be implemented in software, and integrated into the free5GRAN 5G physical layer. The implemented MIMO transmission schemes are Spatial multiplexing based on V-BLAST and Alamouti Space Frequency Block Coding based on LTE Transmission Mode 2. Results show that MIMO decoding adds significant processing latency that may not respect the time budget for decoding. To reduce latency, functions have been implemented using AVX2 instructions and processing times between sequential and AVX2 execution are compared. Measurements are performed in a Faraday cage and the code will be open-source to provide reproductible results. (10.1145/3609382.3610512)
  DOI : 10.1145/3609382.3610512