arXiv subjects

Nan Zhang

Publications and source records attributed to Nan Zhang.

At least 109 records · Page 6Linked to original sources

arXiv2024

Knowledge-driven AI-generated data for accurate and interpretable breast ultrasound diagnoses

Data-driven deep learning models have shown great capabilities to assist radiologists in breast ultrasound (US) diagnoses. However, their effectiveness is limited by the long-tail distribution of training data, which leads to inaccuracies in rare cases. In this study, we address a long-standing challenge of improving the diagnostic model performance on rare cases using long-tailed data. Specifically, we introduce a pipeline, TAILOR, that builds a knowledge-driven generative model to produce tailored synthetic data. The generative model, using 3,749 lesions as source data, can generate millions of breast-US images, especially for error-prone rare cases. The generated data can be further used to build a diagnostic model for accurate and interpretable diagnoses. In the prospective external evaluation, our diagnostic model outperforms the average performance of nine radiologists by 33.5% in specificity with the same sensitivity, improving their performance by providing predictions with an interpretable decision-making process. Moreover, on ductal carcinoma in situ (DCIS), our diagnostic model outperforms all radiologists by a large margin, with only 34 DCIS lesions in the source data. We believe that TAILOR can potentially be extended to various diseases and imaging modalities.

eess.IV↗

arXiv2024

MSCT: Addressing Time-Varying Confounding with Marginal Structural Causal Transformer for Counterfactual Post-Crash Traffic Prediction

Traffic crashes profoundly impede traffic efficiency and pose economic challenges. Accurate prediction of post-crash traffic status provides essential information for evaluating traffic perturbations and developing effective solutions. Previous studies have established a series of deep learning models to predict post-crash traffic conditions, however, these correlation-based methods cannot accommodate the biases caused by time-varying confounders and the heterogeneous effects of crashes. The post-crash traffic prediction model needs to estimate the counterfactual traffic speed response to hypothetical crashes under various conditions, which demonstrates the necessity of understanding the causal relationship between traffic factors. Therefore, this paper presents the Marginal Structural Causal Transformer (MSCT), a novel deep learning model designed for counterfactual post-crash traffic prediction. To address the issue of time-varying confounding bias, MSCT incorporates a structure inspired by Marginal Structural Models and introduces a balanced loss function to facilitate learning of invariant causal features. The proposed model is treatment-aware, with a specific focus on comprehending and predicting traffic speed under hypothetical crash intervention strategies. In the absence of ground-truth data, a synthetic data generation procedure is proposed to emulate the causal mechanism between traffic speed, crashes, and covariates. The model is validated using both synthetic and real-world data, demonstrating that MSCT outperforms state-of-the-art models in multi-step-ahead prediction performance. This study also systematically analyzes the impact of time-varying confounding bias and dataset distribution on model performance, contributing valuable insights into counterfactual prediction for intelligent transportation systems.

cs.LG↗

arXiv2024

Sustainable Wireless Networks via Reconfigurable Intelligent Surfaces (RISs): Overview of the ETSI ISG RIS

Reconfigurable Intelligent Surfaces (RISs) are a novel form of ultra-low power devices that are capable to increase the communication data rates as well as the cell coverage in a cost- and energy-efficient way. This is attributed to their programmable operation that enables them to dynamically manipulate the wireless propagation environment, a feature that has lately inspired numerous research investigations and applications. To pave the way to the formal standardization of RISs, the European Telecommunications Standards Institute (ETSI) launched the Industry Specification Group (ISG) on the RIS technology in September 2021. This article provides a comprehensive overview of the status of the work conducted by the ETSI ISG RIS, covering typical deployment scenarios of reconfigurable metasurfaces, use cases and operating applications, requirements, emerging hardware architectures and operating modes, as well as the latest insights regarding future directions of RISs and the resulting smart wireless environments.

eess.SP↗

arXiv2024

Enhanced Semantic Segmentation Pipeline for WeatherProof Dataset Challenge

This report describes the winning solution to the WeatherProof Dataset Challenge (CVPR 2024 UG2+ Track 3). Details regarding the challenge are available at https://cvpr2024ug2challenge.github.io/track3.html. We propose an enhanced semantic segmentation pipeline for this challenge. Firstly, we improve semantic segmentation models, using backbone pretrained with Depth Anything to improve UperNet model and SETRMLA model, and adding language guidance based on both weather and category information to InternImage model. Secondly, we introduce a new dataset WeatherProofExtra with wider viewing angle and employ data augmentation methods, including adverse weather and super-resolution. Finally, effective training strategies and ensemble method are applied to improve final performance further. Our solution is ranked 1st on the final leaderboard. Code will be available at https://github.com/KaneiGi/WeatherProofChallenge.

cs.CV↗

arXiv2024

Distributed Simulation for Digital Twins of Large-Scale Real-World DiffServ-Based Networks

Digital Twin technology facilitates the monitoring and online analysis of large-scale communication networks. Faster predictions of network performance thus become imperative, especially for analysing Quality of Service (QoS) parameters in large-scale city networks. Discrete Event Simulation (DES) is a standard network analysis technology, and can be further optimised with parallel and distributed execution for speedup, referred to as Parallel Discrete Event Simulation (PDES). However, modelling detailed QoS mechanisms such as DiffServ requires complex event handling for each network router, which can involve excessive simulation events. In addition, current PDES for network analysis mostly adopts conservative scheduling, which suffers from excessive global synchronisation to avoid causality problems. The performance analysis of optimistic PDES for real-world large-scale network topology and complex QoS mechanisms is still inadequate. To address these gaps, this paper proposes a simulation toolkit, Quaint, which leverages an optimistic PDES engine ROSS, for detailed modelling of DiffServ-based networks. A novel event-handling model for each network router is also proposed to significantly reduce the number of events in complex QoS modelling. Quaint has been evaluated using a real-world metropolitan-scale network topology with 5,000 routers/switches. Results show that compared to the conventional simulator OMNeT++/INET, even the sequential mode of Quaint can achieve a speedup of 53 times, and the distributed mode has a speedup of 232 times. Scalability characterisation is conducted to portray the efficiency of distributed execution, and the results indicate the future direction for workload-aware model partitioning.

cs.DC↗

arXiv2024

Compact Spin-Polarized Positron Acceleration in Multi-Layer Microhole Array Films

Compact spin-polarized positron accelerators play a major role in promoting significant positron application research, which typically require high acceleration gradients and polarization degree, both of which, however, are still great challenging. Here, we put forward a novel spin-polarized positron acceleration method which employs an ultrarelativistic high-density electron beam passing through any hole of multi-layer microhole array films to excite strong electrostatic and transition radiation fields. Positrons in the polarized electron-positron pair plasma, filled in the front of the multi-layer films, can be captured, accelerated, and focused by the electrostatic and transition radiation fields, while maintaining high polarization of above 90% and high acceleration gradient of about TeV/m. Multi-layer design allows for capturing more positrons and achieving cascade acceleration. Our method offers a promising solution for accelerator miniaturization, positron injection, and polarization maintaining, and also can be used to accelerate other charged particles.

physics.plasm-ph↗

arXiv2024

Pruning as a Domain-specific LLM Extractor

Large Language Models (LLMs) have exhibited remarkable proficiency across a wide array of NLP tasks. However, the escalation in model size also engenders substantial deployment costs. While few efforts have explored model pruning techniques to reduce the size of LLMs, they mainly center on general or task-specific weights. This leads to suboptimal performance due to lacking specificity on the target domain or generality on different tasks when applied to domain-specific challenges. This work introduces an innovative unstructured dual-pruning methodology, D-Pruner, for domain-specific compression on LLM. It extracts a compressed, domain-specific, and task-agnostic LLM by identifying LLM weights that are pivotal for general capabilities, like linguistic capability and multi-task solving, and domain-specific knowledge. More specifically, we first assess general weight importance by quantifying the error incurred upon their removal with the help of an open-domain calibration dataset. Then, we utilize this general weight importance to refine the training loss, so that it preserves generality when fitting into a specific domain. Moreover, by efficiently approximating weight importance with the refined training loss on a domain-specific calibration dataset, we obtain a pruned model emphasizing generality and specificity. Our comprehensive experiments across various tasks in healthcare and legal domains show the effectiveness of D-Pruner in domain-specific compression. Our code is available at https://github.com/psunlpgroup/D-Pruner.

cs.CL↗

arXiv2024

Fair Abstractive Summarization of Diverse Perspectives

People from different social and demographic groups express diverse perspectives and conflicting opinions on a broad set of topics such as product reviews, healthcare, law, and politics. A fair summary should provide a comprehensive coverage of diverse perspectives without underrepresenting certain groups. However, current work in summarization metrics and Large Language Models (LLMs) evaluation has not explored fair abstractive summarization. In this paper, we systematically investigate fair abstractive summarization for user-generated data. We first formally define fairness in abstractive summarization as not underrepresenting perspectives of any groups of people, and we propose four reference-free automatic metrics by measuring the differences between target and source perspectives. We evaluate nine LLMs, including three GPT models, four LLaMA models, PaLM 2, and Claude, on six datasets collected from social media, online reviews, and recorded transcripts. Experiments show that both the model-generated and the human-written reference summaries suffer from low fairness. We conduct a comprehensive analysis of the common factors influencing fairness and propose three simple but effective methods to alleviate unfair summarization. Our dataset and code are available at https://github.com/psunlpgroup/FairSumm.

cs.CL↗

arXiv2024

PEaCE: A Chemistry-Oriented Dataset for Optical Character Recognition on Scientific Documents

Optical Character Recognition (OCR) is an established task with the objective of identifying the text present in an image. While many off-the-shelf OCR models exist, they are often trained for either scientific (e.g., formulae) or generic printed English text. Extracting text from chemistry publications requires an OCR model that is capable in both realms. Nougat, a recent tool, exhibits strong ability to parse academic documents, but is unable to parse tables in PubMed articles, which comprises a significant part of the academic community and is the focus of this work. To mitigate this gap, we present the Printed English and Chemical Equations (PEaCE) dataset, containing both synthetic and real-world records, and evaluate the efficacy of transformer-based OCR models when trained on this resource. Given that real-world records contain artifacts not present in synthetic records, we propose transformations that mimic such qualities. We perform a suite of experiments to explore the impact of patch size, multi-domain training, and our proposed transformations, ultimately finding that models with a small patch size trained on multiple domains using the proposed transformations yield the best performance. Our dataset and code is available at https://github.com/ZN1010/PEaCE.

cs.CL↗

arXiv2024

Parallel numerical simulation of impact crater with perfect matched layers

Impact craters are the primary geomorphic features on the surfaces of celestial bodies such as the Moon, and their formation has significant implications for the evolutionary history of the celestial body. The study of the impact crater formation process relies mainly on numerical simulation methods, with two-dimensional simulations capable of reproducing general patterns of impact processes while conserving computational resources. However, to mitigate the artificial reflections of shock waves at numerical boundaries, a common approach involves expanding the computational domain, greatly reducing the efficiency of numerical simulations. In this study, we developed a novel two-dimensional code SALEc-2D that employs the perfect matched layer (PML) method to suppress artificial reflections at numerical boundaries. This method enhances computational efficiency while ensuring reliable results. Additionally, we implemented MPI parallel algorithms in the new code to further improve computational efficiency. Simulations that would take over ten hours using the conventional iSALE-2D code can now be completed in less than half an hour using our code, SALEc-2D, on a standard computer. We anticipate that our code will find widespread application in numerical simulations of impact craters in the future.

astro-ph.EP↗

arXiv2024

An efficient method for calculating resonant modes in biperiodic photonic structures

Many photonic devices, such as photonic crystal slabs, cross gratings, and periodic metasurfaces, are biperiodic structures with two independent periodic directions, and are sandwiched between two homogeneous media. Many applications of these devices are closely related to resonance phenomena. Therefore, efficient computation of resonant modes is crucial in device design and structure analysis. Since resonant modes satisfy outgoing radiation conditions, perfectly matched layers (PMLs) are usually used to truncate the unbounded spatial variable perpendicular to the periodic directions. In this paper, we develop an efficient method without using PMLs to calculate resonant modes in biperiodic structures. We reduce the original eigenvalue problem to a small matrix nonlinear eigenvalue problem which is solved by the contour integral method. Numerical examples show that our method is efficient with respect to memory usage and CPU time, free of spurious solutions, and determines degenerate resonant modes without any difficulty.

physics.comp-ph↗

arXiv2024

Top-L Most Influential Community Detection Over Social Networks (Technical Report)

In many real-world applications such as social network analysis and online marketing/advertising, the community detection is a fundamental task to identify communities (subgraphs) in social networks with high structural cohesiveness. While previous works focus on detecting communities alone, they do not consider the collective influences of users in these communities on other user nodes in social networks. Inspired by this, in this paper, we investigate the influence propagation from some seed communities and their influential effects that result in the influenced communities. We propose a novel problem, named Top-L most Influential Community DEtection (TopL-ICDE) over social networks, which aims to retrieve top-L seed communities with the highest influences, having high structural cohesiveness, and containing user-specified query keywords. In order to efficiently tackle the TopL-ICDE problem, we design effective pruning strategies to filter out false alarms of seed communities and propose an effective index mechanism to facilitate efficient Top-L community retrieval. We develop an efficient TopL-ICDE answering algorithm by traversing the index and applying our proposed pruning strategies. We also formulate and tackle a variant of TopL-ICDE, named diversified top-L most influential community detection (DTopL-ICDE), which returns a set of L diversified communities with the highest diversity score (i.e., collaborative influences by L communities). We prove that DTopL-ICDE is NP-hard, and propose an efficient greedy algorithm with our designed diversity score pruning. Through extensive experiments, we verify the efficiency and effectiveness of our proposed TopL-ICDE and DTopL-ICDE approaches over real/synthetic social networks under various parameter settings.

cs.SI↗

arXiv2024

Evidence for Unfolded Fermi Surfaces in the Charge-Density-Wave State of Kagome Metal FeGe Revealed by de Haas-van Alphen Effect

The antiferromagnetic kagome lattice compound FeGe has been revealed to host an emergent charge-density-wave (CDW) state which manifests complex interplay between the spin, charge and lattice degrees of freedom. Here, we present a comprehensive study of the de Haas-van Alphen effect by measuring torque magnetometry under magnetic fields up to 45.2 T to map Fermi surfaces in this unusual CDW state. For field along the $c$ direction, we resolve four cyclotron orbits; the largest one roughly corresponding to the area of the 2$\times$2 folded Brillouin zone. Three smaller orbits are characterized by light effective cyclotron masses range from (0.18-0.30) $m_e$. Angle-resolved measurements identify one Fermi surface segment with weak anisotropy. Combined with band structure calculations, our results suggest that features of unfolded Fermi surfaces are robust against CDW reconstruction, corroborating the novel effect of a short-ranged CDW on the electronic structure.

cond-mat.str-el↗

arXiv2023

The non-perturbative stringy interaction between NS-brane \& Dp brane

To our best knowledge, the leading non-perturbative stringy interaction between an NS brane and a Dp brane remains unknown. We here present the non-perturbative stringy amplitudes for a system of an F-string and a Dp brane and a system of an NS 5 brane and a Dp brane for $0 \le p \le 6$. In either case, the F or NS5 and the Dp are placed parallel at a separation. We obtain the respective amplitudes, starting from the amplitude for a system of a D1 and a D3 for the former and that for a system of a D5 and a D3 system for the latter, based on the IIB S-duality and various T-dualities plus the consistency of both, along with the respective known long-range amplitudes. We would like to point out that the amplitude for the D1/D3 or D3/D5 computed from the usual D-brane technique does not take into consideration of the non-perturbative contribution due to the exchange of virtual closed D-string emitted by the D3. As such the resulting amplitudes obtained from this one via the S-duality and followed by various T-dualities are not consistent with the IIB S-duality. We resolve this issue and obtain the corresponding consistent amplitudes. The implications of so obtained amplitudes are also discussed.

hep-th↗

arXiv2023

Non-generic bound states in the continuum in waveguides with lateral leakage channels

For optical waveguides with a layered background which itself is a slab waveguide, a guided mode is a bound state in the continuum (BIC), if it coexists with slab modes propagating outwards in the lateral direction; i.e., there are lateral leakage channels. It is known that generic BICs in optical waveguides with lateral leakage channels are robust in the sense that they still exist if the waveguide is perturbed arbitrarily. However, the theory is not applicable to non-generic BICs which can be defined precisely. Near a BIC, the waveguide supports resonant and leaky modes with a complex frequency and a complex propagation constant, respectively. In this paper, we develop a perturbation theory to show that the resonant and leaky modes near a non-generic BIC have an ultra-high $Q$ factor and ultra-low leakage loss, respectively. We also show that a merging-BIC obtained by tuning structural parameters is always a non-generic BIC. Existing studies on merging-BICs are concerned with specific examples and specific parameters. We analyze an arbitrary structural perturbation (to a waveguide supporting a non-generic BIC) given by $δF({\bf r})$, where $F({\bf r})$ is the perturbation profile and $δ$ is the amplitude, and show that the perturbed waveguide has two BICs for $δ>0$ (or $δ<0$) and no BIC for $δ<0$ (or $δ>0$). This implies that a non-generic BIC is a merging-BIC (for any perturbation profile $F$) when $δ$ is regarded as a parameter. Our study indicates that non-generic BICs have interesting special properties that are useful in applications.

physics.optics↗

arXiv2023

FaMeSumm: Investigating and Improving Faithfulness of Medical Summarization

Summaries of medical text shall be faithful by being consistent and factual with source inputs, which is an important but understudied topic for safety and efficiency in healthcare. In this paper, we investigate and improve faithfulness in summarization on a broad range of medical summarization tasks. Our investigation reveals that current summarization models often produce unfaithful outputs for medical input text. We then introduce FaMeSumm, a framework to improve faithfulness by fine-tuning pre-trained language models based on medical knowledge. FaMeSumm performs contrastive learning on designed sets of faithful and unfaithful summaries, and it incorporates medical terms and their contexts to encourage faithful generation of medical terms. We conduct comprehensive experiments on three datasets in two languages: health question and radiology report summarization datasets in English, and a patient-doctor dialogue dataset in Chinese. Results demonstrate that FaMeSumm is flexible and effective by delivering consistent improvements over mainstream language models such as BART, T5, mT5, and PEGASUS, yielding state-of-the-art performances on metrics for faithfulness and general quality. Human evaluation by doctors also shows that FaMeSumm generates more faithful outputs. Our code is available at https://github.com/psunlpgroup/FaMeSumm .

cs.CL↗

arXiv2023

Efficient non-collinear antiferromagnetic state switching induced by orbital Hall effect in chromium

Recently orbital Hall current has attracted attention as an alternative method to switch the magnetization of ferromagnets. Here we present our findings on electrical switching of antiferromagnetic state in Mn3Sn/Cr, where despite the much smaller spin Hall angle of Cr, the switching current density is comparable to heavy metal based heterostructures. On the other hand, the inverse process, i.e., spin-to-charge conversion in Cr-based heterostructures is much less efficient than the Pt-based equivalents, as manifested in the almost one order of magnitude smaller terahertz emission intensity and spin current induced magnetoresistance in Cr-based structures. These results in combination with the slow decay of terahertz emission against Cr thickness (diffusion length of ~11 nm) suggest that the observed magnetic switching can be attributed to orbital current generation in Cr, followed by efficient conversion to spin current. Our work demonstrates the potential of light metals like Cr as an efficient orbital/spin current source for antiferromagnetic spintronics.

cond-mat.mtrl-sci↗

arXiv2023

A Digital Twin Approach for Adaptive Compliance in Cyber-Physical Systems: Case of Smart Warehouse Logistics

Engineering regulatory compliance in complex Cyber-Physical Systems (CPS), such as smart warehouse logistics, is challenging due to the open and dynamic nature of these systems, scales, and unpredictable modes of human-robot interactions that can be best learnt at runtime. Traditional offline approaches for engineering compliance often involve modelling at a higher, more abstract level (e.g. using languages like SysML). These abstract models only support analysis in offline-designed and simplified scenarios. However, open and complex systems may be unpredictable, and their behaviours are difficult to be fully captured by abstract models. These systems may also involve other business goals, possibly conflicting with regulatory compliance. To overcome these challenges, fine-grained simulation models are promising to complement abstract models and support accurate runtime predictions and performance evaluation with trade-off analysis. The novel contribution of this work is a Digital Twin-oriented architecture for adaptive compliance leveraging abstract goal modelling, fine-grained agent-based modelling and runtime simulation for managing compliance trade-offs. A case study from smart warehouse logistics is used to demonstrate the approach considering safety and productivity trade-offs.

eess.SY↗