arXiv subjects

Xin Liu

Publications and source records attributed to Xin Liu.

At least 541 records · Page 30Linked to original sources

arXiv2024

The Road to Near-Capacity CV-QKD Reconciliation: An FEC-Agnostic Design

New near-capacity continuous-variable quantum key distribution (CV-QKD) reconciliation schemes are proposed, where both the authenticated classical channel (ClC) and the quantum channel (QuC) for QKD are protected by separate forward error correction (FEC) coding schemes. More explicitly, a new codeword-based - rather than syndrome-based - QKD reconciliation scheme is proposed, where Alice sends an FEC-protected codeword to Bob through a ClC, while Bob sends a separate FEC protected codeword to Alice through a QuC. Upon decoding the codeword received from the other side, the final key is obtained by applying a simple modulo-2 operation to the local codeword and the decoded remote codeword. As a result, first of all, the proposed codeword-based QKD reconciliation system ensures protection of both the QuC and of the ClC. Secondly, the proposed system has a similar complexity at both sides, where both Alice and Bob have an FEC encoder and an FEC decoder. Thirdly, the proposed system makes QKD reconciliation compatible with a wide range of FEC schemes, including polar codes, CCs and irregular convolutional codes (IRCCs), where a near-capacity performance can be achieved for both the QuC and for the ClC.Our simulation results demonstrate that thanks to the proposed regime, the performance improvements of the QuC and of the ClC benefit each other, hence leading to an improved secret key rate (SKR) that inches closer to both the Pirandola-Laurenza-Ottaviani-Banchi (PLOB) bound and to the maximum achieveable rate bound.

quant-ph↗

arXiv2024

Parafermions with symmetry-protected non-Abelian statistics

Non-Abelian anyons have garnered extensive attention for obeying exotic non-Abelian statistics and having potential applications to fault-tolerant quantum computing. While the prior research has predominantly focused on non-Abelian statistics without the necessity of symmetry protection, recent progresses have shown that symmetries can play essential roles and bring a notion of the symmetry-protected non-Abelian (SPNA) statistics. In this work, we extend the concept of SPNA statistics to strongly-correlated systems which host parafermion zero modes (PZMs). This study involves a few fundamental results proved here. First, we unveil a generic unitary symmetry mechanism that protects PZMs from local couplings. Then, with this symmetry protection, the PZMs can be categorized into two nontrivial sectors, each maintaining its own parity conservation, even though the whole system cannot be dismantled into separate subsystems due to nonlinear interactions. Finally, by leveraging the parity conservation of each sector and the general properties of the effective braiding Hamiltonian, we prove rigorously that the PZMs intrinsically obey SPNA statistics. To further confirm the results, we derive the braiding matrix at a tri-junction. We also propose a correlated quantum nanowire model that accommodates a pair of PZMs protected by mirror symmetry and satisfying the generic theory. This work shows a broad spectrum of strongly-correlated systems capable of hosting fractional SPNA quasiparticles and enriches our comprehension of fundamental quantum statistics linked to the symmetries that govern the exchange dynamics.

cond-mat.str-el↗

arXiv2024

Sampling-Based Methods for Multi-Block Optimization Problems over Transport Polytopes

This paper focuses on multi-block optimization problems over transport polytopes, which underlie various applications including strongly correlated quantum physics and machine learning. Conventional block coordinate descent-type methods for the general multi-block problems store and operate on the matrix variables directly, resulting in formidable expenditure for large-scale settings. On the other hand, optimal transport problems, as a special case, have attracted extensive attention and numerical techniques that waive the use of the full matrices have recently emerged. However, it remains nontrivial to apply these techniques to the multi-block, possibly nonconvex problems with theoretical guarantees. In this work, we leverage the benefits of both sides and develop novel sampling-based block coordinate descent-type methods, which are equipped with either entropy regularization or Kullback-Leibler divergence. Each iteration of these methods solves subproblems restricted on the sampled degrees of freedom. Consequently, they involve only sparse matrices, which amounts to considerable complexity reductions. We explicitly characterize the sampling-induced errors and establish convergence and asymptotic properties for the methods equipped with the entropy regularization. Numerical experiments on typical strongly correlated electron systems corroborate their superior scalability over the methods utilizing full matrices. The advantage also enables the first visualization of approximate optimal transport maps between electron positions in three-dimensional contexts.

math.OC↗

arXiv2024

Scalable Path Level Thermal History Simulation of PBF process validated by Melt Pool Images

In this paper we outline the development of a scalable PBF thermal history simulation built on CAPL and based on melt pool physics and dynamics. The new approach inherits linear scalability from CAPL and has three novel ingredients. Firstly, to simulate the laser scanning on a solid surface, we discretize the entire simulation domain instead of only the manufacturing toolpath by appending fictitious paths to the manufacturing toolpath. Secondly, to simulate the scanning on overlapping toolpaths, the path-scale simulations are initialized by a Voronoi diagram for line segments discretized from the manufacturing toolpath. Lastly, we propose a modified conduction model that considers the high thermal gradient around the melt pool. We validate the simulation against melt pool images captured with the co-axial melt pool monitoring (MPM) system on the NIST Additive Manufacturing Metrology Testbed (AMMT). Excellent agreements in the length and width of melt pools are found between simulations and experiments conducted on a custom-controlled laser powder bed fusion (LPBF) testbed on a nickel-alloy (IN625) solid surface. To the authors' best knowledge, this paper is the first to validate a full path-scale thermal history with experimentally acquired melt pool images. Comparing the simulation results and the experimental data, we discuss the influence of laser power on the melt pool length on the path-scale level. We also identify the possible ways to further improve the accuracy of the CAPL simulation without sacrificing efficiency.

cs.CE↗

arXiv2024

Generation is better than Modification: Combating High Class Homophily Variance in Graph Anomaly Detection

Graph-based anomaly detection is currently an important research topic in the field of graph neural networks (GNNs). We find that in graph anomaly detection, the homophily distribution differences between different classes are significantly greater than those in homophilic and heterophilic graphs. For the first time, we introduce a new metric called Class Homophily Variance, which quantitatively describes this phenomenon. To mitigate its impact, we propose a novel GNN model named Homophily Edge Generation Graph Neural Network (HedGe). Previous works typically focused on pruning, selecting or connecting on original relationships, and we refer to these methods as modifications. Different from these works, our method emphasizes generating new relationships with low class homophily variance, using the original relationships as an auxiliary. HedGe samples homophily adjacency matrices from scratch using a self-attention mechanism, and leverages nodes that are relevant in the feature space but not directly connected in the original graph. Additionally, we modify the loss function to punish the generation of unnecessary heterophilic edges by the model. Extensive comparison experiments demonstrate that HedGe achieved the best performance across multiple benchmark datasets, including anomaly detection and edgeless node classification. The proposed model also improves the robustness under the novel Heterophily Attack with increased class homophily variance on other graph classification tasks.

cs.LG↗

arXiv2024

Benchmarking Zero-Shot Robustness of Multimodal Foundation Models: A Pilot Study

Pre-training image representations from the raw text about images enables zero-shot vision transfer to downstream tasks. Through pre-training on millions of samples collected from the internet, multimodal foundation models, such as CLIP, produce state-of-the-art zero-shot results that often reach competitiveness with fully supervised methods without the need for task-specific training. Besides the encouraging performance on classification accuracy, it is reported that these models close the robustness gap by matching the performance of supervised models trained on ImageNet under natural distribution shift. Because robustness is critical to real-world applications, especially safety-critical ones, in this paper, we present a comprehensive evaluation based on a large-scale robustness benchmark covering 7 natural, 3 synthetic distribution shifts, and 11 adversarial attacks. We use CLIP as a pilot study. We show that CLIP leads to a significant robustness drop compared to supervised ImageNet models on our benchmark, especially under synthetic distribution shift and adversarial attacks. Furthermore, data overlap analysis suggests that the observed robustness under natural distribution shifts could be attributed, at least in part, to data overlap. In summary, our evaluation shows a comprehensive evaluation of robustness is necessary; and there is a significant need to improve the robustness of zero-shot multimodal models.

cs.LG↗

arXiv2024

LitCab: Lightweight Language Model Calibration over Short- and Long-form Responses

A model is considered well-calibrated when its probability estimate aligns with the actual likelihood of the output being correct. Calibrating language models (LMs) is crucial, as it plays a vital role in detecting and mitigating hallucinations of LMs as well as building more trustworthy models. However, standard calibration techniques may not be suited for LM calibration. For instance, post-processing methods such as temperature scaling do not reorder the candidate generations. On the other hand, training-based methods require fine-tuning the entire model, which is impractical for LMs of large scale. We present LitCab, a lightweight calibration mechanism consisting of a single linear layer that takes the input text representation and predicts a bias term, which is then added to the LM output logits. LitCab improves model calibration by only adding < 2% of the original model parameters. For evaluation, we construct CaT, a benchmark consisting of eight text generation tasks, covering responses ranging from short phrases to paragraphs. We test LitCab with Llama2-7B, where it improves calibration across all tasks, reducing the average ECE score by as large as 30%. We further conduct a comprehensive evaluation with multiple popular open-sourced LMs from GPT and LLaMA families, yielding the following key findings: (i) Larger models within the same family exhibit better calibration on tasks with short generation tasks, but not necessarily for longer ones. (ii) GPT-family models show superior calibration compared to LLaMA, Llama2, and Vicuna models, despite having much fewer parameters. (iii) Fine-tuning pretrained model (e.g., LLaMA) with samples of limited purpose (e.g., conversations) may lead to worse calibration, highlighting the importance of fine-tuning setups for calibrating LMs.

cs.CL↗

arXiv2024

Answering Diverse Questions via Text Attached with Key Audio-Visual Clues

Audio-visual question answering (AVQA) requires reference to video content and auditory information, followed by correlating the question to predict the most precise answer. Although mining deeper layers of audio-visual information to interact with questions facilitates the multimodal fusion process, the redundancy of audio-visual parameters tends to reduce the generalization of the inference engine to multiple question-answer pairs in a single video. Indeed, the natural heterogeneous relationship between audiovisuals and text makes the perfect fusion challenging, to prevent high-level audio-visual semantics from weakening the network's adaptability to diverse question types, we propose a framework for performing mutual correlation distillation (MCD) to aid question inference. MCD is divided into three main steps: 1) firstly, the residual structure is utilized to enhance the audio-visual soft associations based on self-attention, then key local audio-visual features relevant to the question context are captured hierarchically by shared aggregators and coupled in the form of clues with specific question vectors. 2) Secondly, knowledge distillation is enforced to align audio-visual-text pairs in a shared latent space to narrow the cross-modal semantic gap. 3) And finally, the audio-visual dependencies are decoupled by discarding the decision-level integrations. We evaluate the proposed method on two publicly available datasets containing multiple question-and-answer pairs, i.e., Music-AVQA and AVQA. Experiments show that our method outperforms other state-of-the-art methods, and one interesting finding behind is that removing deep audio-visual features during inference can effectively mitigate overfitting. The source code is released at http://github.com/rikeilong/MCD-forAVQA.

cs.CV↗

arXiv2024

ACT-MNMT Auto-Constriction Turning for Multilingual Neural Machine Translation

Large language model (LLM) has achieved promising performance in multilingual machine translation tasks through zero/few-shot prompts or prompt-tuning. However, due to the mixture of multilingual data during the pre-training of LLM, the LLM-based translation models face the off-target issue in both prompt-based methods, including a series of phenomena, namely instruction misunderstanding, translation with wrong language and over-generation. For this issue, this paper introduces an \textbf{\underline{A}}uto-\textbf{\underline{C}}onstriction \textbf{\underline{T}}urning mechanism for \textbf{\underline{M}}ultilingual \textbf{\underline{N}}eural \textbf{\underline{M}}achine \textbf{\underline{T}}ranslation (\model), which is a novel supervised fine-tuning mechanism and orthogonal to the traditional prompt-based methods. In this method, \model automatically constructs a constrained template in the target side by adding trigger tokens ahead of the ground truth. Furthermore, trigger tokens can be arranged and combined freely to represent different task semantics, and they can be iteratively updated to maximize the label likelihood. Experiments are performed on WMT test sets with multiple metrics, and the experimental results demonstrate that \model achieves substantially improved performance across multiple translation directions and reduce the off-target phenomena in the translation.

cs.CL↗

arXiv2024

VODKA-JWST: A 3.8 kpc dual quasar at cosmic noon in a powerful starburst galaxy with JWST/MIRI IFU

Dual quasars, two active supermassive black holes at galactic scales, represent crucial objects for studying the impact of galaxy mergers and quasar activity on the star formation rate (SFR) within their host galaxies, particularly at cosmic noon when SFR peaks. We present JWST/MIRI mid-infrared integral field spectroscopy of J074922.96+225511.7, a dual quasar with a projected separation of 3.8 kilo-parsec at a redshift $z$ of 2.17. We detect spatially extended [Fe II] 5.34$\rm μ$m and polycyclic aromatic hydrocarbon (PAH) 3.3$μ$m emissions from the star formation activity in its host galaxy. We derive the SFR of 10$^{3.0\pm0.2}$ M$_{\odot}$ yr$^{-1}$ using PAH 3.3$μ$m, which is five times higher than that derived from the cutoff luminosity of the infrared luminosity function for galaxies at $z\sim2$. While the SFR of J0749+2255 agrees with that of star-forming galaxies of comparable stellar mass at the same redshifts, its molecular gas content falls short of expectations based on the molecular Kennicutt-Schmidt law. This discrepancy may result from molecular gas depletion due to the longer elevated stage of star formation, even after the molecular gas reservoir is depleted. We do not observe any quasar-driven outflow that impacts PAH and [Fe II] in the host galaxy based on the spatially resolved maps. From the expected flux in PAH-based star formation, the [Fe II] line likely originates from the star-forming regions in the host galaxy. Our study highlights the stardust nature of J0749+2255, indicating a potential connection between the dual quasar phase and intense star formation activities.

astro-ph.GA↗

arXiv2024

Revisiting Edge Perturbation for Graph Neural Network in Graph Data Augmentation and Attack

Edge perturbation is a basic method to modify graph structures. It can be categorized into two veins based on their effects on the performance of graph neural networks (GNNs), i.e., graph data augmentation and attack. Surprisingly, both veins of edge perturbation methods employ the same operations, yet yield opposite effects on GNNs' accuracy. A distinct boundary between these methods in using edge perturbation has never been clearly defined. Consequently, inappropriate perturbations may lead to undesirable outcomes, necessitating precise adjustments to achieve desired effects. Therefore, questions of ``why edge perturbation has a two-faced effect?'' and ``what makes edge perturbation flexible and effective?'' still remain unanswered. In this paper, we will answer these questions by proposing a unified formulation and establishing a clear boundary between two categories of edge perturbation methods. Specifically, we conduct experiments to elucidate the differences and similarities between these methods and theoretically unify the workflow of these methods by casting it to one optimization problem. Then, we devise Edge Priority Detector (EPD) to generate a novel priority metric, bridging these methods up in the workflow. Experiments show that EPD can make augmentation or attack flexibly and achieve comparable or superior performance to other counterparts with less time overhead.

cs.LG↗

arXiv2024

FedUV: Uniformity and Variance for Heterogeneous Federated Learning

Federated learning is a promising framework to train neural networks with widely distributed data. However, performance degrades heavily with heterogeneously distributed data. Recent work has shown this is due to the final layer of the network being most prone to local bias, some finding success freezing the final layer as an orthogonal classifier. We investigate the training dynamics of the classifier by applying SVD to the weights motivated by the observation that freezing weights results in constant singular values. We find that there are differences when training in IID and non-IID settings. Based on this finding, we introduce two regularization terms for local training to continuously emulate IID settings: (1) variance in the dimension-wise probability distribution of the classifier and (2) hyperspherical uniformity of representations of the encoder. These regularizations promote local models to act as if it were in an IID setting regardless of the local data distribution, thus offsetting proneness to bias while being flexible to the data. On extensive experiments in both label-shift and feature-shift settings, we verify that our method achieves highest performance by a large margin especially in highly non-IID cases in addition to being scalable to larger models and datasets.

cs.LG↗

arXiv2024

A Simple yet Effective Network based on Vision Transformer for Camouflaged Object and Salient Object Detection

Camouflaged object detection (COD) and salient object detection (SOD) are two distinct yet closely-related computer vision tasks widely studied during the past decades. Though sharing the same purpose of segmenting an image into binary foreground and background regions, their distinction lies in the fact that COD focuses on concealed objects hidden in the image, while SOD concentrates on the most prominent objects in the image. Previous works achieved good performance by stacking various hand-designed modules and multi-scale features. However, these carefully-designed complex networks often performed well on one task but not on another. In this work, we propose a simple yet effective network (SENet) based on vision Transformer (ViT), by employing a simple design of an asymmetric ViT-based encoder-decoder structure, we yield competitive results on both tasks, exhibiting greater versatility than meticulously crafted ones. Furthermore, to enhance the Transformer's ability to model local information, which is important for pixel-level binary segmentation tasks, we propose a local information capture module (LICM). We also propose a dynamic weighted loss (DW loss) based on Binary Cross-Entropy (BCE) and Intersection over Union (IoU) loss, which guides the network to pay more attention to those smaller and more difficult-to-find target objects according to their size. Moreover, we explore the issue of joint training of SOD and COD, and propose a preliminary solution to the conflict in joint training, further improving the performance of SOD. Extensive experiments on multiple benchmark datasets demonstrate the effectiveness of our method. The code is available at https://github.com/linuxsino/SENet.

cs.CV↗

arXiv2024

The $B^0 \to J/ψf_0(1370,1500,1710)$ decays: an opportunity for scalar glueball hunting

The scalars $f_0$ closest to 1.5 GeV contain the mesons $f_0(1370)$, $f_0(1500)$ and $f_0(1710)$, and the latter two ones are usually viewed as the potential candidates for the scalar glueballs. In this work, by including the important contributions from the vertex corrections, we study the decays $B^0 \to J/ψf_0$ within the improved perturbative QCD approach and analyze the possible scalar glueball hunting. Together with the two mixing models, namely, $f_0(1500) (f_0(1710))$ being the primary scalar glueball in model I (II), and two classification scenarios, namely, $f_0$ being the $q\bar q$ excited (ground) states in scenario 1 (2), the branching fractions associated with their ratios for $B^0 \to J/ψf_0$ are evaluated comprehensively. The predictions with still large uncertainties in the considered two mixing models are roughly consistent with currently limited data, which indicates that both more rich data and more precise predictions are urgently demanded to figure out the scalar glueball clearly in the future. Moreover, several interesting ratios between the branching fractions of $B^0 \to J/ψf_0(\to π^+ π^-/K^+ K^-)$ and $B^0 \to J/ψρ^0/ϕ(\to π^+ π^-/ K^+ K^-)$ that could help us to understand the nature of scalar $f_0$ are defined and predicted theoretically. These ratios should be examined in future experiments.

hep-ph↗

arXiv2024

Learning to Schedule Online Tasks with Bandit Feedback

Online task scheduling serves an integral role for task-intensive applications in cloud computing and crowdsourcing. Optimal scheduling can enhance system performance, typically measured by the reward-to-cost ratio, under some task arrival distribution. On one hand, both reward and cost are dependent on task context (e.g., evaluation metric) and remain black-box in practice. These render reward and cost hard to model thus unknown before decision making. On the other hand, task arrival behaviors remain sensitive to factors like unpredictable system fluctuation whereby a prior estimation or the conventional assumption of arrival distribution (e.g., Poisson) may fail. This implies another practical yet often neglected challenge, i.e., uncertain task arrival distribution. Towards effective scheduling under a stationary environment with various uncertainties, we propose a double-optimistic learning based Robbins-Monro (DOL-RM) algorithm. Specifically, DOL-RM integrates a learning module that incorporates optimistic estimation for reward-to-cost ratio and a decision module that utilizes the Robbins-Monro method to implicitly learn task arrival distribution while making scheduling decisions. Theoretically, DOL-RM achieves convergence gap and no regret learning with a sub-linear regret of $O(T^{3/4})$, which is the first result for online task scheduling under uncertain task arrival distribution and unknown reward and cost. Our numerical results in a synthetic experiment and a real-world application demonstrate the effectiveness of DOL-RM in achieving the best cumulative reward-to-cost ratio compared with other state-of-the-art baselines.

cs.LG↗

arXiv2024

OncoGPT: A Medical Conversational Model Tailored with Oncology Domain Expertise on a Large Language Model Meta-AI (LLaMA)

In the past year, there has been a growing trend in applying Large Language Models (LLMs) to the field of medicine, particularly with the advent of advanced language models such as ChatGPT developed by OpenAI. However, there is limited research on LLMs specifically addressing oncology-related queries. The primary aim of this research was to develop a specialized language model that demonstrates improved accuracy in providing advice related to oncology. We performed an extensive data collection of online question-answer interactions centered around oncology, sourced from reputable doctor-patient platforms. Following data cleaning and anonymization, a dataset comprising over 180K+ oncology-related conversations was established. The conversations were categorized and meticulously reviewed by field specialists and clinicians to ensure precision. Employing the LLaMA model and other selected open-source datasets, we conducted iterative fine-tuning to enhance the model's proficiency in basic medical conversation and specialized oncology knowledge. We observed a substantial enhancement in the model's understanding of genuine patient inquiries and its reliability in offering oncology-related advice through the utilization of real online question-answer interactions in the fine-tuning process. We release database and models to the research community (https://github.com/OncoGPT1).

cs.CL↗

arXiv2024

MegaScale: Scaling Large Language Model Training to More Than 10,000 GPUs

We present the design, implementation and engineering experience in building and deploying MegaScale, a production system for training large language models (LLMs) at the scale of more than 10,000 GPUs. Training LLMs at this scale brings unprecedented challenges to training efficiency and stability. We take a full-stack approach that co-designs the algorithmic and system components across model block and optimizer design, computation and communication overlapping, operator optimization, data pipeline, and network performance tuning. Maintaining high efficiency throughout the training process (i.e., stability) is an important consideration in production given the long extent of LLM training jobs. Many hard stability issues only emerge at large scale, and in-depth observability is the key to address them. We develop a set of diagnosis tools to monitor system components and events deep in the stack, identify root causes, and derive effective techniques to achieve fault tolerance and mitigate stragglers. MegaScale achieves 55.2% Model FLOPs Utilization (MFU) when training a 175B LLM model on 12,288 GPUs, improving the MFU by 1.34x compared to Megatron-LM. We share our operational experience in identifying and fixing failures and stragglers. We hope by articulating the problems and sharing our experience from a systems perspective, this work can inspire future LLM systems research.

cs.LG↗

arXiv2024

Demographic Bias of Expert-Level Vision-Language Foundation Models in Medical Imaging

Advances in artificial intelligence (AI) have achieved expert-level performance in medical imaging applications. Notably, self-supervised vision-language foundation models can detect a broad spectrum of pathologies without relying on explicit training annotations. However, it is crucial to ensure that these AI models do not mirror or amplify human biases, thereby disadvantaging historically marginalized groups such as females or Black patients. The manifestation of such biases could systematically delay essential medical care for certain patient subgroups. In this study, we investigate the algorithmic fairness of state-of-the-art vision-language foundation models in chest X-ray diagnosis across five globally-sourced datasets. Our findings reveal that compared to board-certified radiologists, these foundation models consistently underdiagnose marginalized groups, with even higher rates seen in intersectional subgroups, such as Black female patients. Such demographic biases present over a wide range of pathologies and demographic attributes. Further analysis of the model embedding uncovers its significant encoding of demographic information. Deploying AI systems with these biases in medical imaging can intensify pre-existing care disparities, posing potential challenges to equitable healthcare access and raising ethical questions about their clinical application.

cs.CY↗