Search

Your search keyword '"Yao, Yifan"' showing total 453 results
Start Over Author "Yao, Yifan"
453 results on '"Yao, Yifan"'

1. Ramsey and Gallai-Ramsey numbers for linear forests and kipas

Author

Li, Ping, Mao, Yaping, Schiermeyer, Ingo, and Yao, Yifan

Subjects
Mathematics - Combinatorics
Abstract

For two graphs $G,H$, the \emph{Ramsey number} $r(G,H)$ is the minimum integer $n$ such that any red/blue edge-coloring of $K_n$ contains either a red copy of $G$ or a blue copy of $H$. For two graphs $G,H$, the \emph{Gallai-Ramsey number} $\operatorname{gr}_k(G:H)$ is defined as the minimum integer $n$ such that any $k$-edge-coloring of $K_n$ must contain either a rainbow copy of $G$ or a monochromatic copy of $H$. In this paper, the classical Ramsey numbers of linear forest versus kipas are obtained. We obtain the exact values of $\operatorname{gr}_k(G:H)$, where $H$ is either a path or a kipas and $G\in\{K_{1,3},P_4^+,P_5\}$ and $P_4^+$ is the graph consisting of $P_4$ with one extra edge incident with inner vertex.

Published
2024

2. A Survey on Large Language Model (LLM) Security and Privacy: The Good, the Bad, and the Ugly

Author

Yao, Yifan, Duan, Jinhao, Xu, Kaidi, Cai, Yuanfang, Sun, Zhibo, and Zhang, Yue

Subjects
Computer Science - Cryptography and Security, Computer Science - Artificial Intelligence
Abstract

Large Language Models (LLMs), such as ChatGPT and Bard, have revolutionized natural language understanding and generation. They possess deep language comprehension, human-like text generation capabilities, contextual awareness, and robust problem-solving skills, making them invaluable in various domains (e.g., search engines, customer support, translation). In the meantime, LLMs have also gained traction in the security community, revealing security vulnerabilities and showcasing their potential in security-related tasks. This paper explores the intersection of LLMs with security and privacy. Specifically, we investigate how LLMs positively impact security and privacy, potential risks and threats associated with their use, and inherent vulnerabilities within LLMs. Through a comprehensive literature review, the paper categorizes the papers into "The Good" (beneficial LLM applications), "The Bad" (offensive applications), and "The Ugly" (vulnerabilities of LLMs and their defenses). We have some interesting findings. For example, LLMs have proven to enhance code security (code vulnerability detection) and data privacy (data confidentiality protection), outperforming traditional methods. However, they can also be harnessed for various attacks (particularly user-level attacks) due to their human-like reasoning abilities. We have identified areas that require further research efforts. For example, Research on model and parameter extraction attacks is limited and often theoretical, hindered by LLM parameter scale and confidentiality. Safe instruction tuning, a recent development, requires more exploration. We hope that our work can shed light on the LLMs' potential to both bolster and jeopardize cybersecurity.

Published
2023
Full Text
View/download PDF

3. Non-equilibrium dynamics of electron emission from cold and hot graphene under proton irradiation

Author

Yao, Yifan, Kononov, Alina, Metzlaff, Arne, Wucher, Andreas, Kalkhoff, Lukas, Breuer, Lars, Schleberger, Marika, and Schleife, André

Subjects
Condensed Matter - Materials Science
Abstract

Characteristic properties of secondary electrons emitted from irradiated two-dimensional materials arise from multi-length and time-scale relaxation processes that connect the initial non-equilibrium excited electron distribution with their eventual emission. To understand these processes, which are critical for using secondary electrons as high-resolution thermalization probes, we combine first-principles real-time electron dynamics with modern experiments. Our data for cold and hot proton-irradiated graphene shows signatures of kinetic and potential emission and generally good agreement for electron yields between experiment and theory. The duration of the emission pulse is about 1.5 femtoseconds, indicating high time resolution when used as a probe. Our newly developed method to predict kinetic energy spectra shows good agreement with electron and ion irradiation experiments and prior models. We find that lattice temperature significantly increases secondary electron emission, whereas electron temperature has a negligible effect.

Published
2023

4. Towards Few-shot Out-of-Distribution Detection

Author

Dong, Jiuqing, Gao, Yongbin, Zhou, Heng, Cen, Jun, Yao, Yifan, Yoon, Sook, and Sun, Park Dong

Subjects
Computer Science - Computer Vision and Pattern Recognition
Abstract

Out-of-distribution (OOD) detection is critical for ensuring the reliability of open-world intelligent systems. Despite the notable advancements in existing OOD detection methodologies, our study identifies a significant performance drop under the scarcity of training samples. In this context, we introduce a novel few-shot OOD detection benchmark, carefully constructed to address this gap. Our empirical analysis reveals the superiority of ParameterEfficient Fine-Tuning (PEFT) strategies, such as visual prompt tuning and visual adapter tuning, over conventional techniques, including fully fine-tuning and linear probing tuning in the few-shot OOD detection task. Recognizing some crucial information from the pre-trained model, which is pivotal for OOD detection, may be lost during the fine-tuning process, we propose a method termed DomainSpecific and General Knowledge Fusion (DSGF). This approach is designed to be compatible with diverse fine-tuning frameworks. Our experiments show that the integration of DSGF significantly enhances the few-shot OOD detection capabilities across various methods and fine-tuning methodologies, including fully fine-tuning, visual adapter tuning, and visual prompt tuning. The code will be released.

Published
2023

5. IDHPre: Intradialytic Hypotension Prediction Model Based on Fully Observed Features

Author

Yao, Yifan, Kuang, Zemin, Yang, Xiwen, Wang, Baoquan, Niu, Zhaomeng, Yang, Jiaxin, Hu, Lun, Zhou, Xi, Hu, Pengwei, Goos, Gerhard, Series Editor, Hartmanis, Juris, Founding Editor, Bertino, Elisa, Editorial Board Member, Gao, Wen, Editorial Board Member, Steffen, Bernhard, Editorial Board Member, Yung, Moti, Editorial Board Member, Huang, De-Shuang, editor, Zhang, Qinhu, editor, and Guo, Jiayang, editor

Published
2024
Full Text
View/download PDF

6. Electron dynamics in extended systems within real-time time-dependent density functional theory

Author

Kononov, Alina, Lee, Cheng-Wei, Santos, Tatiane Pereira dos, Robinson, Brian, Yao, Yifan, Yao, Yi, Andrade, Xavier, Baczewski, Andrew David, Constantinescu, Emil, Correa, Alfredo A., Kanai, Yosuke, Modine, Normand, and Schleife, Andre

Subjects
Condensed Matter - Materials Science, Physics - Computational Physics
Abstract

Due to a beneficial balance of computational cost and accuracy, real-time time-dependent density functional theory has emerged as a promising first-principles framework to describe electron real-time dynamics. Here we discuss recent implementations around this approach, in particular in the context of complex, extended systems. Results include an analysis of the computational cost associated with numerical propagation and when using absorbing boundary conditions. We extensively explore the shortcomings for describing electron-electron scattering in real time and compare to many-body perturbation theory. Modern improvements of the description of exchange and correlation are reviewed. In this work, we specifically focus on the Qb@ll code, which we have mainly used for these types of simulations over the last years, and we conclude by pointing to further progress needed going forward.

Published
2022

12. Cervical cancer screening aided by artificial intelligence, China/Depistage du cancer du col de l'uterus assiste par intelligence artificielle en Chine/Deteccion del cancer del cuello uterino con ayuda de la inteligencia artificial en China

Author

Zhu, Xingce, Yao, Qiang, Dai, Wei, Ji, Lu, Yao, Yifan, Pang, Baochuan, Turic, Bojana, Yao, Lan, and Liu, Zhiyong

Subjects
Artificial intelligence, Cancer -- Diagnosis, Cervical cancer, Artificial intelligence, Health, World Health Organization -- Evaluation
Abstract

Objective To implement and evaluate a large-scale online cervical cancer screening programme in Hubei Province, China, supported by artificial intelligence and delivered by trained health workers. Methods The screening programme, which started in 2017, used four types of health worker: sampling health workers, slide preparation technicians, diagnostic health workers and cytopathologists. Sampling health workers took samples from the women on site; slide preparation technicians prepared slides for liquid-based cytology; diagnostic health workers identified negative samples and classified positive samples based on the Bethesda System after cytological assessment using online artificial intelligence; and cytopathologists reviewed positive samples and signed reports of the results online. The programme used fully automated scanners, online artificial intelligence, an online screening management platform, and mobile telephone devices to provide screening services. We evaluated the sustainability, performance and cost of the programme. Results From 2017 to 2021, 1 518 972 women in 16 cities in Hubei Province participated in the programme, of whom 1 474 788 (97.09%) had valid samples for the screening. Of the 86 648 women whose samples were positive, 30 486 required a biopsy but only 19 495 had one. The biopsy showed that 2785 women had precancerous lesions and 191 had invasive cancers. The cost of screening was 6.31 United States dollars (US$) per woman for the public payer: US$ 1.03 administrative costs and US$ 5.28 online screening costs. Conclusion Cervical cancer screening using artificial intelligence in Hubei Province provided a low-cost, accessible and effective service, which will contribute to achieving universal cervical cancer screening coverage in China. Objectif Deployer et evaluer un programme en ligne de grande envergure destine au depistage du cancer du col de l'uterus dans la province du Hubei, en Chine, reposant sur l'intelligence artificielle et mis en oeuvre par des agents de sante formes a cet effet. Methodes Le programme de depistage, qui a debute en 2017, s'appuyait sur quatre types d'agents de sante: des agents charges de prelever les echantillons, des techniciens charges de preparer les lames, des agents charges d'etablir le diagnostic et des cytopathologistes. Les premiers ont preleve des echantillons aupres des femmes sur place, les deuxiemes ont prepare les lames pour la cytologie en phase liquide, les troisiemes ont identifie les echantillons negatifs et classe les echantillons positifs selon le systeme de Bethesda apres analyse cytologique a l'aide d'une intelligence artificielle en ligne, tandis que les cytopathologistes ont examine les echantillons positifs et signe les rapports de resultats en ligne. Le programme etait constitue de scanners entierement automatises, d'une intelligence artificielle en ligne, d'une plateforme de gestion du depistage en ligne, ainsi que d'appareils de telephonie mobile pour les services de depistage. Nous avons evalue la viabilite, les performances et le cout d'un tel programme. Resultats Entre 2017 et 2021, 1 518 972 femmes residant dans 16 villes de la province du Hubei ont participe au programme, et 1 474 788 (97,09%) d'entre elles ont presente des echantillons valides pour le depistage. Sur les 86 648 femmes dont les echantillons se sont reveles positifs, 30 486 ont eu besoin d'une biopsie, mais seulement 19 495 en ont effectue une. La biopsie a montre que 2785 femmes presentaient des lesions precancereuses et que 191 etaient atteintes de cancers invasifs. Le cout du depistage s'elevait a 6,31 dollars americains (USD) par femme pour le payeur public: 1,03 USD pour les frais administratifs et 5,28 USD pour le depistage en ligne. Conclusion Le depistage du cancer du col de l'uterus assiste par intelligence artificielle dans la province du Hubei est un service abordable, accessible et efficace, qui contribuera a garantir une couverture universelle en la matiere dans le pays. Objetivo Implementar y evaluar un programa de deteccion del cancer del cuello uterino en linea a gran escala en la provincia de Hubei (China), con el apoyo de la inteligencia artificial y a cargo de trabajadores sanitarios formados. Metodos El programa de deteccion, que comenzo en 2017, utilizo cuatro tipos de trabajadores sanitarios: trabajadores sanitarios de muestreo, tecnicos de preparacion de portaobjetos, trabajadores sanitarios de diagnostico y citopatologos. Los trabajadores sanitarios de muestreo tomaban muestras de las mujeres in situ, los tecnicos de preparacion de portaobjetos preparaban los portaobjetos para la citologia mediante liquidos, los trabajadores sanitarios de diagnostico identificaban las muestras negativas y clasificaban las muestras positivas segun el Sistema Bethesda tras la evaluacion citologica mediante inteligencia artificial en linea, y los citopatologos revisaban las muestras positivas y firmaban los informes de los resultados en linea. El programa utilizo escaneres totalmente automatizados, inteligencia artificial en linea, una plataforma de gestion de la deteccion en linea y dispositivos de telefonia movil para proporcionar servicios de deteccion. Se evaluaron la sostenibilidad, el rendimiento y el coste del programa. Resultados Entre 2017 y 2021, 1 518 972 mujeres de 16 ciudades de la provincia de Hubei participaron en el programa, de las que 1 474 788 (97,09%) tenian muestras validas para la deteccion. De las 86 648 mujeres cuyas muestras dieron positivo, 30 486 necesitaron una biopsia, pero solo 19 495 se sometieron a ella. La biopsia mostro que 2785 mujeres tenian lesiones precancerosas y 191 canceres invasivos. El coste de la deteccion fue de 6,31 dolares estadounidenses (US$) por mujer para el organismo publico pagador: US$ 1,03 de costes administrativos y US$ 5,28 de costes de deteccion en linea. Conclusion La deteccion del cancer del cuello uterino mediante inteligencia artificial en la provincia de Hubei proporciono un servicio de bajo coste, accesible y eficaz, que contribuira a lograr la cobertura universal de deteccion del cancer del cuello uterino en China., Introduction Cervical cancer is a threat to women's health. According to data released by the International Agency for Research on Cancer, the cancer research agency of the World Health Organization [...]

Published
2023
Full Text
View/download PDF

20. PanGu-$\alpha$: Large-scale Autoregressive Pretrained Chinese Language Models with Auto-parallel Computation

Author

Zeng, Wei, Ren, Xiaozhe, Su, Teng, Wang, Hui, Liao, Yi, Wang, Zhiwei, Jiang, Xin, Yang, ZhenZhang, Wang, Kaisheng, Zhang, Xiaoda, Li, Chen, Gong, Ziyan, Yao, Yifan, Huang, Xinjing, Wang, Jun, Yu, Jianfeng, Guo, Qi, Yu, Yue, Zhang, Yan, Wang, Jin, Tao, Hengtao, Yan, Dasen, Yi, Zexuan, Peng, Fang, Jiang, Fangqing, Zhang, Han, Deng, Lingfeng, Zhang, Yehong, Lin, Zhe, Zhang, Chao, Zhang, Shaojie, Guo, Mingyue, Gu, Shanzhi, Fan, Gaojun, Wang, Yaowei, Jin, Xuefeng, Liu, Qun, and Tian, Yonghong

Subjects
Computer Science - Computation and Language
Abstract

Large-scale Pretrained Language Models (PLMs) have become the new paradigm for Natural Language Processing (NLP). PLMs with hundreds of billions parameters such as GPT-3 have demonstrated strong performances on natural language understanding and generation with \textit{few-shot in-context} learning. In this work, we present our practice on training large-scale autoregressive language models named PanGu-$\alpha$, with up to 200 billion parameters. PanGu-$\alpha$ is developed under the MindSpore and trained on a cluster of 2048 Ascend 910 AI processors. The training parallelism strategy is implemented based on MindSpore Auto-parallel, which composes five parallelism dimensions to scale the training task to 2048 processors efficiently, including data parallelism, op-level model parallelism, pipeline model parallelism, optimizer model parallelism and rematerialization. To enhance the generalization ability of PanGu-$\alpha$, we collect 1.1TB high-quality Chinese data from a wide range of domains to pretrain the model. We empirically test the generation ability of PanGu-$\alpha$ in various scenarios including text summarization, question answering, dialogue generation, etc. Moreover, we investigate the effect of model scales on the few-shot performances across a broad range of Chinese NLP tasks. The experimental results demonstrate the superior capabilities of PanGu-$\alpha$ in performing various tasks under few-shot or zero-shot settings., Comment: The technique report for PanGu-$\alpha$

Published
2021

21. Learning Transferable Kinematic Dictionary for 3D Human Pose and Shape Reconstruction

Author

Ma, Ze, Yao, Yifan, Ji, Pan, and Ma, Chao

Subjects
Computer Science - Computer Vision and Pattern Recognition
Abstract

Estimating 3D human pose and shape from a single image is highly under-constrained. To address this ambiguity, we propose a novel prior, namely kinematic dictionary, which explicitly regularizes the solution space of relative 3D rotations of human joints in the kinematic tree. Integrated with a statistical human model and a deep neural network, our method achieves end-to-end 3D reconstruction without the need of using any shape annotations during the training of neural networks. The kinematic dictionary bridges the gap between in-the-wild images and 3D datasets, and thus facilitates end-to-end training across all types of datasets. The proposed method achieves competitive results on large-scale datasets including Human3.6M, MPI-INF-3DHP, and LSP, while running in real-time given the human bounding boxes.

Published
2021

24. Combination of window-sliding and prediction range method based on LSTM model for predicting cryptocurrency

Author

Yao, Yifan and Wang, Lina

Subjects
Quantitative Finance - Statistical Finance
Abstract

The present study aims to establish the model of the cryptocurrency price trend based on financial theory using the LSTM model with multiple combinations between the window length and the predicting horizons, the random walk model is also applied with different parameter settings., Comment: 30 pages,23 figures

Published
2021

37. Development of high efficiency AlGaN-based UV LEDs on Sapphire

Author

Yao, Yifan

Subjects
Materials Science, AlGaN, GaN, MOCVD, Polarization Engineering, UV LEDs
Abstract

AlGaN-based ultraviolet LEDs attract significant interests recently for their versatile applications. Germicidal irradiation in the UV-C wavelengths can combat future pandemic outbreaks while wavelengths in the longer UV-B/UV-A range have wideapplications in biochemical sensing and medical treatment. However, commercial AlGaN-based ultraviolet LEDs have much lower power efficiency (< 5%) than the traditional incumbent technologies mercury lamps due to many challenges this ultra-wide bandgap material poses in the epitaxial MOCVD growth, device design and fabrication technologies. In this dissertation, I will first discuss AlN epitaxy technologies on SiC and sapphire substrates by MOCVD. With optimized growth condition, we have achieved highly conductive n-type Al0.6Ga0.4N with around 5 mΩ-cm and demonstrated low contact resistance Vanadium based ohmic contact on n-type Al0.6Ga0.4N. I will then present how doping efficiency in p-type AlGaN can be massively improved with polarization engineering to achieve efficient carrier injection for UV LEDs. Next, I will discuss some aspects of the active region device design and MOCVD growth optimization that can further enhance the device performance. Combining the developments in each of these areas with improved fabrication and packagingtechnologies, we have demonstrated UV LEDs with world class efficiencies from deep UV (below 250 nm) to UV-A wavelength (320+ nm). Lastly, I will present some recent developments of UV-A LEDs results on GaN templates and discuss how some of these technologies can be utilized for future UV laser diodes developments.

Published
2023

39. Comparing Discrete and Empirical Troposphere Delay Models: A Global IGS‐Based Evaluation.

Author

Yao, Yifan, Yang, Fei, Li, Jian, Wang, Lei, Zheng, Junxi, Hao, Ruixian, and Xu, Tairan

Subjects
GLOBAL Positioning System, TROPOSPHERE, ROOT-mean-squares, TROPOSPHERIC chemistry
Abstract

Zenith tropospheric delay (ZTD) is an important atmospheric parameter in radio‐space‐geodetic techniques such as Global Navigation Satellite System (GNSS), which is pivotal for GNSS positioning, navigation and meteorology. The Vienna Mapping Function (VMF) data server is a widely utilized source for implementing ZTD, offering two types of models, that is, the empirical one and the discrete one with Grid‐wise and Site‐wise models. Therefore, to evaluate the accuracy of these models becomes the focus of this article. Specifically, this study investigates their performances in terms of calculation of ZTD, using the hourly values derived from the International GNSS Service data as references. The results show that the root mean square err (RMSE) of the Site‐wise, Grid‐wise and global pressure and temperature 3 model are 11.71/13.03/38.56 mm, respectively, indicating the discrete model performs generally better than the empirical model, and the Site‐wise model is the better of the two discrete models. From the perspective of spatial resolution, the performance of these three models in ZTD calculation shows obvious influences of latitude changes and elevation differences. From the temporal analysis, the accuracy of the discrete model shows differences over different UTC epochs, while the empirical model can only express the seasonal ZTD characteristics with the average RMSE at different epochs being similar, the specifically values are 39.67, 39.26, 39.38 and 39.18 mm at UTC 0:00, 6:00, 12:00 and 18:00, respectively. The histogram and boxplot well indicate the accuracy differences of the three models in different seasons. Additionally, the time series of three models at different latitudes were also explored in this research. These explorations are conducive to the selection of appropriate models for calculating ZTD based on specific requirements. Key Points: The discrete and empirical troposphere delay models of the VMF data server is comprehensibly evaluated using the global International GNSS Service (IGS) dataThe Site‐wise of discrete model exhibits the best accuracy and is recommended first, followed by the Grid‐wide and the empirical modelEach model shows different influences by the latitude, elevation, seasons and epochs [ABSTRACT FROM AUTHOR]

Published
2024
Full Text
View/download PDF

47. Tight Toughness, Isolated Toughness and Binding Number Bounds for the Star-Path Factor.

Author

Yao, Yifan, Liang, Jinxia, Li, Yue, and Ma, Rongrong

Subjects
*GRAPH connectivity
Abstract

For a set of connected graphs, a spanning subgraph H of G is called a -factor if each component of H is isomorphic to a member of . In this paper, some sufficient conditions with regard to tight toughness, isolated toughness and binding number bounds to guarantee the existence of the { K 1 , j : 1 ≤ j ≤ 2 k } -factor and { P 2 , P 5 } -factor for any graph are obtained. [ABSTRACT FROM AUTHOR]

Published
2024
Full Text
View/download PDF

50. Monitoring-edge-geodetic numbers of radix triangular mesh and Sierpiński graphs.

Author

Ma, Rongrong, Ji, Zhen, Yao, Yifan, and Lei, Yalong

Subjects
MESH networks, GASKETS
Abstract

Given a graph G and an edge $ e\in E(G) $ e ∈ E (G) , let S be a vertex set of G. For any two vertices x, $ y\in S $ y ∈ S , if e belongs to all the shortest paths between x and y, then x and y can monitor the edge e. For each edge e of G, if there exists x and y in S such that x and y can monitor e, then the set S can be called a monitoring-edge-geodetic ( $ \operatorname {MEG} $ MEG for short) set of G. The $ \operatorname {MEG} $ MEG number, denoted by $ \operatorname {meg}(G) $ meg ⁡ (G) , is the size of the smallest $ \operatorname {MEG} $ MEG set of G. In this paper, we obtain the exact values of the $ \operatorname {MEG} $ MEG numbers for radix triangular mesh networks, Sierpiński graphs, Sierpiński gasket graphs and Sierpiński generalized graphs. [ABSTRACT FROM AUTHOR]

Published
2024
Full Text
View/download PDF

Catalog

Books, media, physical & digital resources
See catalog results