检索结果-内蒙古大学图书馆

Toward the Understanding of Deep Text Matching Models for Information Retrieval

学校读者我要写书评

暂无评论

arXiv 2021年

作者： Chen, Lijuan Lan, Yanyan Pang, Liang Guo, Jiafeng Cheng, Xueqi Sogou Inc. Beijing China Institute for AI Industry Research Tsinghua University Beijing China Data Intelligence System Research Center Institute of Computing Technology Chinese Academy of Sciences Beijing China CAS Key Lab of Network Data Science and Technology Institute of Computing Technology Chinese Academy of Sciences Beijing China

Semantic text matching is a critical problem in information retrieval. Recently, deep learning techniques have been widely used in this area and obtained significant performance improvements. However, most models are black boxes and it is hard to understand what happened in the matching process, due to the poor interpretability of deep learning. This paper aims at tackling this problem. The key idea is to test whether existing deep text matching methods satisfy some fundamental heuristics in information retrieval. Specifically, four heuristics are used in our study, i.e., term frequency constraint, term discrimination constraint, length normalization constraints, and TF-length constraint. Since deep matching models usually contain many parameters, it is difficult to conduct a theoretical study for these complicated functions. In this paper, We propose an empirical testing method. Specifically, We first construct some queries and documents to make them satisfy the assumption in a constraint, and then test to which extend a deep text matching model trained on the original dataset satisfy the corresponding constraint. Besides, a famous attribution based interpretation method, namely integrated gradient, is adopted to conduct detailed analysis and guide for feasible improvement. Experimental results on LETOR 4.0 and MS Marco show that all the investigated deep text matching methods, both representation and interaction based methods, satisfy the above constraints with high probabilities in statistics. We further extend these constraints to the semantic settings, which are shown to be better satisfied for all the deep text matching models. These empirical findings give clear understandings on why deep text matching models usually perform well in information retrieval. We believe the proposed evaluation methodology will be useful for testing future deep text matching models. © 2021, CC BY.

关键词： Deep learning

Proposal for stable atom trapping on a GaN-on-Sapphire chip

学校读者我要写书评

暂无评论

arXiv 2022年

作者： Liu, Aiping Xu, Lei Xu, Xin-Biao Chen, Guang-Jie Zhang, Pengfei Xiang, Guo-Yong Guo, Guang-Can Wang, Qin Zou, Chang-Ling Institute of quantum information and technology Nanjing University of Posts and Telecommunications Nanjing210003 China CAS Key Laboratory of Quantum Information University of Science and Technology of China Hefei230026 China State Key Laboratory of Quantum Optics and Quantum Optics Devices Institute of Opto-Electronics Shanxi University Taiyuan030006 China Broadband wireless communication and sensor network technology Key lab of Ministry of Education Nanjing University of Posts and Telecommunications Nanjing210003 China CAS Center for Excellence in Quantum Information and Quantum Physics University of Science and Technology of China Hefei230026 China

The hybrid photon-atom integrated circuits, which include photonic microcavities and trapped single neutral atom in their evanescent field, are of great potential for quantum information processing. In this platform, the atoms provide the single-photon nonlinearity and long-lived memory, which are complementary to the excellent passive photonics devices in conventional quantum photonic circuits. In this work, we propose a stable platform for realizing the hybrid photon-atom circuits based on an unsuspended photonic chip. By introducing high-order modes in the microring, a feasible evanescent-field trap potential well ∼ 0.3mK could be obtained by only 10mW-level power in the cavity, compared with 100mW-level power required in the scheme based on fundamental modes. Based on our scheme, stable single atom trapping with relatively low laser power is feasible for future studies on high-fidelity quantum gates, single-photon sources, as well as many-body quantum physics based on a controllable atom array in a microcavity. © 2022, CC BY.

关键词： Atoms

Multi-grating design for integrated single-atom trapping, manipulation and readout

学校读者我要写书评

暂无评论

arXiv 2022年

作者： Liu, Aiping Liu, Jiawei Peng, Wei Xu, Xin-Biao Chen, Guang-Jie Ren, Xifeng Wang, Qin Zou, Chang-Ling Institute of Quantum Information and Technology Nanjing University of Posts and Telecommunications Nanjing210003 China CAS Key Laboratory of Quantum Information University of Science and Technology of China Hefei230026 China Broadband Wireless Communication and Sensor Network Technology Key Lab of Ministry of Education Nanjing University of Posts and Telecommunications Nanjing210003 China CAS Center for Excellence in Quantum Information and Quantum Physics University of Science and Technology of China Hefei230026 China State Key Laboratory of Quantum Optics and Quantum Optics Devices Institute of Opto-Electronics Shanxi University Taiyuan030006 China

An on-chip multi-grating device is proposed to interface single-atoms and integrated photonic circuits, by guiding and focusing lasers to the area with ∼10 μm above the chip for trapping, state manipulation and readout of single Rubidium atoms. For the optical dipole trap, two 850 nm laser beams are diffracted and overlapped to form a lattice of single-atom dipole trap, with the diameter of optical dipole trap around 2.7 μm. Similar gratings are designed for guiding 780 nm probe laser to excite and also collect the fluorescence of 87Rb atoms. Such a device provides a compact solution for future applications of single atoms, including the single photon source, single-atom quantum register, and sensor. © 2022, CC BY.

关键词： Atoms

Self-supervised GANs with label augmentation 21

学校读者我要写书评

暂无评论

Self-supervised GANs with label augmentation

Proceedings of the 35th International Conference on Neural Information Processing Systems

作者： Liang Hou Huawei Shen Qi Cao Xueqi Cheng Data Intelligence System Research Center Institute of Computing Technology Chinese Academy of Sciences and University of Chinese Academy of Sciences Data Intelligence System Research Center Institute of Computing Technology Chinese Academy of Sciences CAS Key Laboratory of Network Data Science and Technology Institute of Computing Technology Chinese Academy of Sciences and University of Chinese Academy of Sciences

ISBN: (纸本)9781713845393

Recently, transformation-based self-supervised learning has been applied to generative adversarial networks (GANs) to mitigate catastrophic forgetting in the discriminator by introducing a stationary learning environment. However, the separate self-supervised tasks in existing self-supervised GANs cause a goal inconsistent with generative modeling due to the fact that their self-supervised classifiers are agnostic to the generator distribution. To address this problem, we propose a novel self-supervised GAN that unifies the GAN task with the self-supervised task by augmenting the GAN labels (real or fake) via self-supervision of data transformation. Specifically, the original discriminator and self-supervised classifier are unified into a label-augmented discriminator that predicts the augmented labels to be aware of both the generator distribution and the data distribution under every transformation, and then provide the discrepancy between them to optimize the generator. Theoretically, we prove that the optimal generator could converge to replicate the real data distribution. Empirically, we show that the proposed method significantly outperforms previous self-supervised and data augmentation GANs on both generative modeling and representation learning across benchmark datasets.

关键词：

GCN-ALP: Addressing Matching Collisions in Anchor Link Prediction

学校读者我要写书评

暂无评论

GCN-ALP: Addressing Matching Collisions in Anchor Link Predi...

IEEE International Conference on Big Knowledge (ICBK)

作者： Hao Gao Yongqing Wang Shanshan Lyu Huawei Shen Xueqi Cheng CAS Key Laboratory of Network Data Science and Technology Institute of Computing Technology Chinese Academy of Sciences Beijing China

ISBN: (数字)9781728181561

ISBN: (纸本)9781728181578

Nowadays online users prefer to join multiple social media for the purpose of socialized online service. The problem anchor link prediction is formalized to link user data with the common ground on user profile, content and network structure across social networks. Most of the traditional works concentrated on learning matching function with explicit or implicit features on observed user data. However, the low quality of observed user data confuses the judgment on anchor links, resulting in the matching collision problem in practice. In this paper, we explore local structure consistency and then construct a matching graph in order to circumvent matching collisions. Furthermore, we propose graph convolution networks with mini-batch strategy, efficiently solving anchor link prediction on matching graph. The experimental results on three real application scenarios show the great potentials of our proposed method in both prediction accuracy and efficiency. In addition, the visualization of learned embeddings provides us a qualitative way to understand the inference of anchor links on the matching graph.

关键词： Convolution Feature extraction Flickr Couplings Computer architecture Trajectory

Toward Enhanced Robustness in Unsupervised Graph Representation Learning: A Graph Information Bottleneck Perspective

学校读者我要写书评

暂无评论

arXiv 2022年

作者： Wang, Jihong Luo, Minnan Li, Jundong Liu, Ziqi Zhou, Jun Zheng, Qinghua The Ministry of Education Key Lab for Intelligent Networks and Network Security School of Computer Science and Technology Xi’an Jiaotong University Xi’an710049 China The Department of Electrical and Computer Engineering Department of Computer Science School of Data Science University of Virginia United States The Ant Financial Services Group Zhejiang Hangzhou310000 China

Recent studies have revealed that GNNs are vulnerable to adversarial attacks. Most existing robust graph learning methods measure model robustness based on label information, rendering them infeasible when label information is not available. A straightforward direction is to employ the widely used Infomax technique from typical Unsupervised Graph Representation Learning (UGRL) to learn robust unsupervised representations. Nonetheless, directly transplanting the Infomax technique from typical UGRL to robust UGRL may involve a biased assumption. In light of the limitation of Infomax, we propose a novel unbiased robust UGRL method called Robust Graph Information Bottleneck (RGIB), which is grounded in the Information Bottleneck (IB) principle. Our RGIB attempts to learn robust node representations against adversarial perturbations by preserving the original information in the benign graph while eliminating the adversarial information in the adversarial graph. There are mainly two challenges to optimize RGIB: 1) high complexity of adversarial attack to perturb node features and graph structure jointly in the training procedure;2) mutual information estimation upon adversarially attacked graphs. To tackle these problems, we further propose an efficient adversarial training strategy with only feature perturbations and an effective mutual information estimator with subgraph-level summary. Moreover, we theoretically establish a connection between our proposed RGIB and the robustness of downstream classifiers, revealing that RGIB can provide a lower bound on the adversarial risk of downstream classifiers. Extensive experiments over several benchmarks and downstream tasks demonstrate the effectiveness and superiority of our proposed method. Copyright © 2022, The Authors. All rights reserved.

关键词： Learning systems

Exploration of the Influence on Training Deep Learning Models by Watermarked Image dataset

学校读者我要写书评

暂无评论

Exploration of the Influence on Training Deep Learning Model...

IEEE International Conference on Big data and Cloud Computing (BdCloud)

作者： Shiqin Liu Shiyuan Feng Jinxia Wu Wei Ren Weiqi Wang Wenwen Zheng School of Computer Science China University of Geosciences Wuhan China Key Laboratory of Network Assessment Technology CAS Chinese Academy of Sciences Beijing China Guizhou Provincial Key Laboratory of Public Big Data Guizhou University Guiyang China

Deep learning has achieved great success in various applications with the help of a large scale of datasets. As a result, sharing the valuable big data that can be applied to training deep learning models is of essential importance currently. However, how to claim ownership and protect the copyright of image big data during the sharing process is still a vital issue that should be addressed. The application of digital watermarks can protect the copyright of image data, at the same time, it also degrades the image quality at the same time. As for invisible digital watermarks, the higher the watermark embedding intensity causes the greater the host image changes. Therefore, the performance of deep learning models may decline due to using the watermarked training set. In this paper, we evaluate the influences of different embedding intensities of various watermarking algorithms on several mainstream models and conclude how the watermarking intensity affects the model training. Besides, referring to watermarking algorithms that have been proposed, we proposed a novel discrete Fourier transform-based watermarking algorithm that can achieve image copyright protection yet maintain the utility of models.

关键词： Training Deep learning Image quality Fourier transforms Watermarking Big data Copyright protection

ICTNET at Trec 2019 Decision Track 28

学校读者我要写书评

暂无评论

ICTNET at Trec 2019 Decision Track

28th Text REtrieval Conference, TREC 2019

作者： Cui, Wanqing Jiang, Yan Tao, Shuchang Guo, Hanzhang CAS Key Lab of Network Data Science and Technology Institute of Computing Technology Chinese Academy of Sciences China University of Chinese Academy of Sciences China

In this paper we report on our participation in the Trec 2019 Decision Track[1] which aims to provide a venue for research on retrieval methods that promote better decision making with search engines and develop new online and offline evaluation methods to predict the decision making quality induced by search results. We convert this task into a standard information retrieval task and use traditional IR model. Finally we give a summary for the solution of our work. © 2019 28th Text REtrieval Conference, TREC 2019 - Proceedings. All Rights Reserved.

关键词： Decision making

Improving the Automated Coronal Jet Identification with U-NET

学校读者我要写书评

暂无评论

arXiv 2024年

作者： Liu, Jiajia Ji, Chunyu Wang, Yimin Soós, Szabolcs Jiang, Ye Erdélyi, Robertus Korsós, M.B. Wang, Yuming Deep Space Exploration Lab School of Earth and Space Sciences University of Science and Technology of China Hefei230026 China CAS Key Laboratory of Geospace Environment CAS Center for Excellence in Comparative Planetology Mengcheng National Geophysical Observatory University of Science and Technology of China Hefei230026 China School of Data Science Qingdao University of Science and Technology Qingdao266100 China Department of Astronomy Eötvös Loránd University Pázmány P. sétány 1/A BudapestH-1117 Hungary Petőfi tér 3. GyulaH-5700 Hungary School of Information Science and Technology Qingdao University of Science and Technology Qingdao266100 China School of Mathematics and Statistics The University of Sheffield SheffieldS3 7RH United Kingdom University of Sheffield Department of Automatic Control and Systems Engineering Amy Johnson Building Portabello Street SheffieldS1 3JD United Kingdom

Coronal jets are one of the most common eruptive activities in the solar atmosphere. They are related to rich physics processes, including but not limited to magnetic reconnection, flaring, instabilities, and plasma heating. Automated identification of off-limb coronal jets has been difficult due to their abundant nature, complex appearance, and relatively small size compared to other features in the corona. In this paper, we present an automated coronal jet identification algorithm (AJIA) that utilizes true and fake jets previously detected by a laborious semi-automated jet detection algorithm (SAJIA, Liu et al. 2023) as the input of an image segmentation neural network U-NET. It is found that AJIA could achieve a much higher (0.81) detecting precision than SAJIA (0.34), meanwhile giving the possibility of whether each pixel in an input image belongs to a jet. We demonstrate that with the aid of artificial neural networks, AJIA could enable fast, accurate, and real-time coronal jet identification from SDO/AIA 304 Å observations, which are essential in studying the collective and long-term behavior of coronal jets and their relation with the solar activity cycles. Copyright © 2024, The Authors. All rights reserved.

关键词： Image segmentation