Search

Emerging 5G/6G use cases span various industries, necessitating flexible solutions that leverage emerging technologies to meet diverse and stringent application requirements under changing network conditions. The standard 5G RAN solution, retransmission, reduces packet loss but can increase transmission delay in the process. Random Linear Network Coding (RLNC) offers an alternative by proactively sending combinations of original packets, thus reducing both delay and packet loss. Current research often only simulates the integration of RLNC in 5G while we implement and evaluate our approach on real commercially available hardware in a real-world deployment. We introduce Flexible Network Coding (FlexNC), which enables the flexible fusion of several RLNC protocols by incorporating a forwarder with multiple RLNC nodes. Network operators can configure FlexNC based on network conditions and application requirements. To further boost network programmability, our Recoder in the Network (RecNet) leverages intermediate network nodes to join the coding process. Both the proposed algorithms have been implemented on OpenAirInterface and extensively tested with traffic from different applications in a real network. While FlexNC adapts to various application needs of latency and packet loss, RecNet significantly minimizes packet loss for a remote user with minimal increase in delay compared to pure RLNC.

A spin-photon interface is one of the key components of a quantum network. Physical platforms under investigation span the range of modern experimental physics, from ultra-cold atoms and ions to a variety of solid-state systems. Each system has its strengths and weaknesses, typically with a trade-off between spin properties and photonic properties. Currently, the best deterministic single-photon sources use a semiconductor quantum dot embedded in an optical microcavity. However, coherent spin control has not yet been integrated with a state-of-the-art single-photon source, and the magnetic noise from host nuclear spins in the semiconductor environment has placed strong limitations on the spin coherence. Here, we combine high-fidelity all-optical spin control with a quantum dot in an open microcavity, currently the most efficient single-photon source platform available. By imprinting a microwave signal onto a red-detuned optical field, a Raman process, we demonstrate coherent rotations of a hole spin around an arbitrary axis of the Bloch sphere, achieving a maximum {\pi}-pulse fidelity of 98.6%. The cavity enhances the Raman process, enabling ultra-fast Rabi frequencies above 1 GHz. We use our flexible spin control to perform optical cooling of the nuclear spins in the host material via the central hole spin, extending the hole-spin coherence time T2* from 28 ns to 535 ns. Hahn echo preserves the spin coherence on a timescale of 20 {\mu}s, and dynamical decoupling extends the coherence close to the relaxation limit. Both the spin T2* and spin rotation time are much larger than the Purcell-enhanced radiative recombination time, 50 ps, enabling many spin-photon pairs to be created before the spin loses its coherence., Comment: 15 pages, 3 main figures, 6 supplementary figures

Via thousands of papers in Explainable AI (XAI), attention maps \cite{vaswani2017attention} and feature importance maps \cite{bansal2020sam} have been established as a common means for finding how important each input feature is to an AI's decisions. It is an interesting, unexplored question whether allowing users to edit the feature importance at test time would improve a human-AI team's accuracy on downstream tasks. In this paper, we address this question by leveraging CHM-Corr, a state-of-the-art, ante-hoc explainable classifier \cite{taesiri2022visual} that first predicts patch-wise correspondences between the input and training-set images, and then bases on them to make classification decisions. We build CHM-Corr++, an interactive interface for CHM-Corr, enabling users to edit the feature importance map provided by CHM-Corr and observe updated model decisions. Via CHM-Corr++, users can gain insights into if, when, and how the model changes its outputs, improving their understanding beyond static explanations. However, our study with 18 expert users who performed 1,400 decisions finds no statistical significance that our interactive approach improves user accuracy on CUB-200 bird image classification over static explanations. This challenges the hypothesis that interactivity can boost human-AI team accuracy and raises needs for future research. We open-source CHM-Corr++, an interactive tool for editing image classifier attention (see an interactive demo here: http://137.184.82.109:7080/). We release code and data on github: https://github.com/anguyen8/chm-corr-interactive., Comment: Accepted for presentation at the XAI4CV Workshop, part of the CVPR 2024 proceedings

We present experiments in which self-assembled InAs quantum dots are coupled to a thin, suspended-beam GaAs resonator. The quantum dots are driven resonantly and the resonance fluorescence is detected. The narrow quantum-dot linewidths, just a factor of three larger than the transform limit, result in a high sensitivity to the mechanical motion. We show that one quantum dot couples to eight mechanical modes spanning a frequency range from $30$ to $600~\mathrm{MHz}$: one quantum dot provides an extensive characterisation of the mechanical resonator. The coupling spans the unresolved-sideband to the resolved-sideband regimes. Finally, we present the first detection of thermally-driven phonon sidebands (at $4.2~\mathrm{K}$) in the resonance-fluoresence spectrum.

A promising route towards the heralded creation and annihilation of single-phonons is to couple a single-photon emitter to a mechanical resonator. The challenge lies in reaching the resolved-sideband regime with a large coupling rate and a high mechanical quality factor. We achieve all of this by coupling self-assembled InAs quantum dots to a small-mode-volume phononic-crystal resonator with mechanical frequency $\Omega_\mathrm{m}/2\pi = 1.466~\mathrm{GHz}$ and quality factor $Q_\mathrm{m} = 2.1\times10^3$. Thanks to the high coupling rate of $g_\mathrm{ep}/2\pi = 2.9~\mathrm{MHz}$, and by exploiting a matching condition between the effective Rabi and mechanical frequencies, we are able to observe the interaction between the two systems. Our results represent a major step towards quantum control of the mechanical resonator via a single-photon emitter.

Occlusion is a long-standing problem in computer vision, particularly in instance segmentation. ACM MMSports 2023 DeepSportRadar has introduced a dataset that focuses on segmenting human subjects within a basketball context and a specialized evaluation metric for occlusion scenarios. Given the modest size of the dataset and the highly deformable nature of the objects to be segmented, this challenge demands the application of robust data augmentation techniques and wisely-chosen deep learning architectures. Our work (ranked 1st in the competition) first proposes a novel data augmentation technique, capable of generating more training samples with wider distribution. Then, we adopt a new architecture - Hybrid Task Cascade (HTC) framework with CBNetV2 as backbone and MaskIoU head to improve segmentation performance. Furthermore, we employ a Stochastic Weight Averaging (SWA) training strategy to improve the model's generalization. As a result, we achieve a remarkable occlusion score (OM) of 0.533 on the challenge dataset, securing the top-1 position on the leaderboard. Source code is available at this https://github.com/nguyendinhson-kaist/MMSports23-Seg-AutoID.

Robots performing human-scale manipulation tasks require an extensive amount of knowledge about their surroundings in order to perform their actions competently and human-like. In this work, we investigate the use of virtual reality technology as an implementation for robot environment modeling, and present a technique for translating scene graphs into knowledge bases. To this end, we take advantage of the Universal Scene Description (USD) format which is an emerging standard for the authoring, visualization and simulation of complex environments. We investigate the conversion of USD-based environment models into Knowledge Graph (KG) representations that facilitate semantic querying and integration with additional knowledge sources., Comment: 6 pages, 3 figures, ICRA 2024

A spin-photon interface should operate with both coherent photons and a coherent spin to enable cluster-state generation and entanglement distribution. In high-quality devices, self-assembled GaAs quantum dots are near-perfect emitters of on-demand coherent photons. However, the spin rapidly decoheres via the magnetic noise arising from the host nuclei. Here, we address this drawback by implementing an all-optical nuclear-spin cooling scheme on a GaAs quantum dot. The electron-spin coherence time increases 156-fold from $T_2^*$ = 3.9 ns to 0.608 $\mu$s. The cooling scheme depends on a non-collinear term in the hyperfine interaction. The results show that such a term is present even though the strain is low and no external stress is applied. Our work highlights the potential of optically-active GaAs quantum dots as fast, highly coherent spin-photon interfaces.

This paper provides an overview of the hardware and software components used in our test bed project the NET Playground. All source information is stored in the GitLab repository (https://gitlab.com/Paulteck/net-playground). In the Hardware section, we present sketches and 3D views of mechanical parts and technical drawings of printed boards. The Software section discusses relay control using shell scripts and the utilization of Ansible for automation. We also introduce a C++ framework for connecting with the INA231 energy sensor. This paper serves as a reference for understanding and replicating our project's hardware and software components.

Machine learning (ML) is increasingly being used in critical decision-making software, but incidents have raised questions about the fairness of ML predictions. To address this issue, new tools and methods are needed to mitigate bias in ML-based software. Previous studies have proposed bias mitigation algorithms that only work in specific situations and often result in a loss of accuracy. Our proposed solution is a novel approach that utilizes automated machine learning (AutoML) techniques to mitigate bias. Our approach includes two key innovations: a novel optimization function and a fairness-aware search space. By improving the default optimization function of AutoML and incorporating fairness objectives, we are able to mitigate bias with little to no loss of accuracy. Additionally, we propose a fairness-aware search space pruning method for AutoML to reduce computational cost and repair time. Our approach, built on the state-of-the-art Auto-Sklearn tool, is designed to reduce bias in real-world scenarios. In order to demonstrate the effectiveness of our approach, we evaluated our approach on four fairness problems and 16 different ML models, and our results show a significant improvement over the baseline and existing bias mitigation techniques. Our approach, Fair-AutoML, successfully repaired 60 out of 64 buggy cases, while existing bias mitigation techniques only repaired up to 44 out of 64 cases., Comment: In Proceedings of The 31st ACM Joint European Software Engineering Conference and Symposium on the Foundations of Software Engineering (ESEC/FSE 2023)

This paper investigates the convergence of Wong--Zakai approximations to regime-switching stochastic differential equations, generated by a collection of finite-variation approximations to Brownian motion. We extend the results of Nguyen and Peralta (2021) to $\mathbb{R}^d$-valued RSSDE by utilising rough path theoretic tools, acquiring the same modification of rate., Comment: 22 pages, 1 figure

Image classifiers are information-discarding machines, by design. Yet, how these models discard information remains mysterious. We hypothesize that one way for image classifiers to reach high accuracy is to first zoom to the most discriminative region in the image and then extract features from there to predict image labels, discarding the rest of the image. Studying six popular networks ranging from AlexNet to CLIP, we find that proper framing of the input image can lead to the correct classification of 98.91% of ImageNet images. Furthermore, we uncover positional biases in various datasets, especially a strong center bias in two popular datasets: ImageNet-A and ObjectNet. Finally, leveraging our insights into the potential of zooming, we propose a test-time augmentation (TTA) technique that improves classification accuracy by forcing models to explicitly perform zoom-in operations before making predictions. Our method is more interpretable, accurate, and faster than MEMO, a state-of-the-art (SOTA) TTA method. We introduce ImageNet-Hard, a new benchmark that challenges SOTA classifiers including large vision-language models even when optimal zooming is allowed., Comment: NeurIPS 2023 Track on Datasets and Benchmarks

Many applications involve humans in the loop, where continuous and accurate human motion monitoring provides valuable information for safe and intuitive human-machine interaction. Portable devices such as inertial measurement units (IMUs) are applicable to monitor human motions, while in practice often limited computational power is available locally. The human motion in task space coordinates requires not only the human joint motion but also the nonlinear coordinate transformation depending on the parameters such as human limb length. In most applications, measuring these kinematics parameters for each individual requires undesirably high effort. Therefore, it is desirable to estimate both, the human motion and kinematic parameters from IMUs. In this work, we propose a novel computational framework for dual estimation in real-time exploiting in-network computational resources. We adopt the concept of field Kalman filtering, where the dual estimation problem is decomposed into a fast state estimation process and a computationally expensive parameter estimation process. In order to further accelerate the convergence, the parameter estimation is progressively computed on multiple networked computational nodes. The superiority of our proposed method is demonstrated by a simulation of a human arm, where the estimation accuracy is shown to converge faster than with conventional approaches.

The advent of deep learning has led to a significant gain in machine translation. However, most of the studies required a large parallel dataset which is scarce and expensive to construct and even unavailable for some languages. This paper presents a simple yet effective method to tackle this problem for low-resource languages by augmenting high-quality sentence pairs and training NMT models in a semi-supervised manner. Specifically, our approach combines the cross-entropy loss for supervised learning with KL Divergence for unsupervised fashion given pseudo and augmented target sentences derived from the model. We also introduce a SentenceBERT-based filter to enhance the quality of augmenting data by retaining semantically similar sentence pairs. Experimental results show that our approach significantly improves NMT baselines, especially on low-resource datasets with 0.46--2.03 BLEU scores. We also demonstrate that using unsupervised training for augmented data is more efficient than reusing the ground-truth target sentences for supervised learning., Comment: TMP and GN contributed equally

In this paper, we focus on studying the distortion of the exterior conformal modulus of a quadrilateral of sufficiently arbitrary form under the stretching map along the abscissa axis with coefficient $H\to\infty$. By using the properties of quasiconformal transformations and taking into account some facts from the theory of elliptic integrals, we confirm that the asymptotic behavior of this modulus does not depend on the shape of the boundary of the quadrilateral. Especially, it is equivalent to $(1/\pi)\log H$ as $H\to\infty$. Therefore, we give a solution to the Vuorinen problem for the exterior region of an arbitrary quadrilateral., Comment: 12 pages

In this work, we study the distortion of the exterior conformal modulus of a symmetric quadrilateral, when stretched in the direction of the abscissa axis with the coefficient $H\to \infty$. By using some facts from the theory of elliptic integrals, we confirm that the asymptotic behavior of this modulus does not depend on the shape of the boundary of the quadrilateral; moreover, it is equivalent to $(1/\pi)\log H$ as $H\to \infty$., Comment: 13 pages

Malware authors often use cryptographic tools such as XOR encryption and block ciphers like AES to obfuscate part of the malware to evade detection. Use of cryptography may give the impression that these obfuscation techniques have some provable guarantees of success. In this paper, we take a closer look at the use of cryptographic tools to obfuscate malware. We first find that most techniques are easy to defeat (in principle), since the decryption algorithm and the key is shipped within the program. In order to clearly define an obfuscation technique's potential to evade detection we propose a principled definition of malware obfuscation, and then categorize instances of malware obfuscation that use cryptographic tools into those which evade detection and those which are detectable. We find that schemes that are hard to de-obfuscate necessarily rely on a construct based on environmental keying. We also show that cryptographic notions of obfuscation, e.g., indistinghuishability and virtual black box obfuscation, may not guarantee evasion detection under our model. However, they can be used in conjunction with environmental keying to produce hard to de-obfuscate version of programs., Comment: This is the full version of the paper with the same title to appear in the Journal of Computer Virology and Hacking Techniques

Explaining the behaviors of deep neural networks, usually considered as black boxes, is critical especially when they are now being adopted over diverse aspects of human life. Taking the advantages of interpretable machine learning (interpretable ML), this work proposes a novel tool called Catastrophic Forgetting Dissector (or CFD) to explain catastrophic forgetting in continual learning settings. We also introduce a new method called Critical Freezing based on the observations of our tool. Experiments on ResNet articulate how catastrophic forgetting happens, particularly showing which components of this famous network are forgetting. Our new continual learning algorithm defeats various recent techniques by a significant margin, proving the capability of the investigation. Critical freezing not only attacks catastrophic forgetting but also exposes explainability., Comment: Master of Science Thesis at KAIST; 24 pages; Keywords: continual learning, catastrophic forgetting, XAI, attribution map, interpretability

Robust, reliable, and deterministic networks are essential for a variety of applications. In order to provide guaranteed communication network services, Time-Sensitive Networking (TSN) unites a set of standards for time-synchronization, flow control, enhanced reliability, and management. We design the TSN-FlexTest testbed with generic commodity hardware and open-source software components to enable flexible TSN measurements. We have conducted extensive measurements to validate the TSN-FlexTest testbed and to examine TSN characteristics. The measurements provide insights into the effects of TSN configurations, such as increasing the number of synchronization messages for the Precision Time Protocol, indicating that a measurement accuracy of 15 ns can be achieved. The TSN measurements included extensive evaluations of the Time-aware Shaper (TAS) for sets of Tactile Internet (TI) packet traffic streams. The measurements elucidate the effects of different scheduling and shaping approaches, while revealing the need for pervasive network control that synchronizes the sending nodes with the network switches. We present the first measurements of distributed TAS with synchronized senders on a commodity hardware testbed, demonstrating the same Quality-of-Service as with dedicated wires for high-priority TI streams despite a 200% over-saturation cross traffic load. The testbed is provided as an open-source project to facilitate future TSN research., Comment: 30 pages, 18 figures, 6 tables, IEEE TNSM, in print, 2024. Shorter version in print in IEEE Trans. on Network and Service Management (see related DOI below)

The optical spectrum of a quantum dot is typically dominated by the fundamental transition between the lowest-energy configurations. However, the radiative Auger process can result in additional red-shifted emission lines. The origin of these lines is a combination of Coulomb interaction and symmetry-breaking in the quantum dot. In this paper, we present measurements of such radiative Auger lines for a range of InGaAs/GaAs self-assembled quantum dots and use a tight-binding model with a configuration interaction approach to explain their appearance. Introducing a composition fluctuation cluster in the dot, our calculations show excellent agreement with measurements. We relate our findings to group theory explaining the origin of the additional emission lines. Our model and results give insight into the interplay between the symmetry breaking in a quantum dot and the position and strength of the radiative Auger lines.

We describe the asymptotic behavior of the conformal modulus of an unbounded doubly-connected domain, non-symmetric with respect to the coordinate axes, when stretched in the direction of the abscissa axis with coefficient $H\to +\infty$. Therefore, we give a partial answer to a problem, suggested by M.~Vourinen, for the case of an arbitrary unbounded domain., Comment: 15 pages, 5 figures