Filters








977 Hits in 2.5 sec

Weakly Supervised Action Selection Learning in Video [article]

Junwei Ma, Satya Krishna Gorti, Maksims Volkovs, Guangwei Yu
2021 arXiv   pre-print
Localizing actions in video is a core task in computer vision. The weakly supervised temporal localization problem investigates whether this task can be adequately solved with only video-level labels, significantly reducing the amount of expensive and error-prone annotation that is required. A common approach is to train a frame-level classifier where frames with the highest class probability are selected to make a video-level prediction. Frame level activations are then used for localization.
more » ... owever, the absence of frame-level annotations cause the classifier to impart class bias on every frame. To address this, we propose the Action Selection Learning (ASL) approach to capture the general concept of action, a property we refer to as "actionness". Under ASL, the model is trained with a novel class-agnostic task to predict which frames will be selected by the classifier. Empirically, we show that ASL outperforms leading baselines on two popular benchmarks THUMOS-14 and ActivityNet-1.2, with 10.3% and 5.7% relative improvement respectively. We further analyze the properties of ASL and demonstrate the importance of actionness. Full code for this work is available here: https://github.com/layer6ai-labs/ASL.
arXiv:2105.02439v1 fatcat:2x3lr5pu3fbvle62ltee5p4esu

Inter-Subject Analysis: Inferring Sparse Interactions with Dense Intra-Graphs [article]

Cong Ma, Junwei Lu, Han Liu
2017 arXiv   pre-print
We develop a new modeling framework for Inter-Subject Analysis (ISA). The goal of ISA is to explore the dependency structure between different subjects with the intra-subject dependency as nuisance. It has important applications in neuroscience to explore the functional connectivity between brain regions under natural stimuli. Our framework is based on the Gaussian graphical models, under which ISA can be converted to the problem of estimation and inference of the inter-subject precision
more » ... The main statistical challenge is that we do not impose sparsity constraint on the whole precision matrix and we only assume the inter-subject part is sparse. For estimation, we propose to estimate an alternative parameter to get around the non-sparse issue and it can achieve asymptotic consistency even if the intra-subject dependency is dense. For inference, we propose an "untangle and chord" procedure to de-bias our estimator. It is valid without the sparsity assumption on the inverse Hessian of the log-likelihood function. This inferential method is general and can be applied to many other statistical problems, thus it is of independent theoretical interest. Numerical experiments on both simulated and brain imaging data validate our methods and theory.
arXiv:1709.07036v1 fatcat:g6hk7xuvnbbpdcjwxaucsqipze

Cross-Class Relevance Learning for Temporal Concept Localization [article]

Junwei Ma, Satya Krishna Gorti, Maksims Volkovs, Ilya Stanevich, Guangwei Yu
2019 arXiv   pre-print
We present a novel Cross-Class Relevance Learning approach for the task of temporal concept localization. Most localization architectures rely on feature extraction layers followed by a classification layer which outputs class probabilities for each segment. However, in many real-world applications classes can exhibit complex relationships that are difficult to model with this architecture. In contrast, we propose to incorporate target class and class-related features as input, and learn a
more » ... ise binary model to predict general segment to class relevance. This facilitates learning of shared information between classes, and allows for arbitrary class-specific feature engineering. We apply this approach to the 3rd YouTube-8M Video Understanding Challenge together with other leading models, and achieve first place out of over 280 teams. In this paper we describe our approach and show some empirical results.
arXiv:1911.08548v1 fatcat:dot2ofwk5bhpvpftdwiimsimme

Inter-Subject Analysis: A Partial Gaussian Graphical Model Approach

Cong Ma, Junwei Lu, Han Liu
2020 figshare.com  
Different from traditional intra-subject analysis, the goal of Inter-Subject Analysis (ISA) is to explore the dependency structure between different subjects with the intra-subject dependency as nuisance. ISA has important applications in neuroscience to study the functional connectivity between brain regions under natural stimuli. We propose a modeling framework for ISA that is based on Gaussian graphical models, under which ISA can be converted to the problem of estimation and inference of a
more » ... artial Gaussian graphical model. The main statistical challenge is that we do not impose sparsity constraints on the whole precision matrix and we only assume the inter-subject part is sparse. For estimation, we propose to estimate an alternative parameter to get around the non-sparse issue and it can achieve asymptotic consistency even if the intra-subject dependency is dense. For inference, we propose an "untangle and chord" procedure to de-bias our estimator. It is valid without the sparsity assumption on the inverse Hessian of the log-likelihood function. This inferential method is general and can be applied to many other statistical problems, thus it is of independent theoretical interest. Numerical experiments on both simulated and brain imaging data validate our methods and theory.
doi:10.6084/m9.figshare.13157833.v1 fatcat:glnz2e6sk5ee7e6jsrmovem23e

Adversarial Attack on Graph Neural Networks as An Influence Maximization Problem [article]

Jiaqi Ma, Junwei Deng, Qiaozhu Mei
2021 arXiv   pre-print
In particular, Ma et al.  ...  C Additional Experiments C.1 Results under the Experiment Setup by Ma et al. [19] In a previous version of this paper, we conducted experiments under a setup exactly the same as that by Ma et al.  ... 
arXiv:2106.10785v1 fatcat:jtivdltkkbchtoo42da34zwnze

Impact of Environmental Regulation on High-Quality Economic Development

Xiaowei Ma, Junwei Xu
2022 Frontiers in Environmental Science  
Ma and Xu High-Quality Economic Development Variables Nationwide Eastern region Central region Western region (1) (2) (3) (4) (5) (6) (7) (8) hqe lngdp hqe Lngdp hqe lngdp hqe lngdp reg2 −1.133*** −1.251  ... 
doi:10.3389/fenvs.2022.896892 fatcat:6a4vppui7jfqjmxsn3ozrpgq7a

Subgroup Generalization and Fairness of Graph Neural Networks [article]

Jiaqi Ma, Junwei Deng, Qiaozhu Mei
2021 arXiv   pre-print
Minjie Wang, Da Zheng, Zihao Ye, Quan Gan, Mufei Li, Xiang Song, Jinjing Zhou, Chao Ma, Lingfan Yu, Yu Gai, et al.  ... 
arXiv:2106.15535v2 fatcat:zxlgget6vzekxkghwkqjlbc4aq

Molecular binding behavior of bipyridium derivatives by water-soluble carboxylato-biphen[3]arene

Junwei Ma, Hongmei Deng, Suxiang Ma, Jian Li, Xueshun Jia, Chunju Li
2015 Chemical Communications  
Complexation of paraquat derivatives and bis(N-mono-substituted bipyridine) cations by a negatively charged carboxylato-biphen[3]arene in water is described.
doi:10.1039/c5cc01470k pmid:25778925 fatcat:wzl5zpqmlrbw7j4e6m67mv4ffi

Biphen[n]arenes

Huanqing Chen, Jiazeng Fan, Xiaoshi Hu, Junwei Ma, Shilu Wang, Jian Li, Yihua Yu, Xueshun Jia, Chunju Li
2015 Chemical Science  
We describe the one-pot synthesis, unique geometries and intriguing host–guest properties of a new class of supramolecular macrocycles – biphen[n]arenes (n = 3, 4).
doi:10.1039/c4sc02422b pmid:28553468 pmcid:PMC5424672 fatcat:4jhrpgqihngqba2e7svpr4zmmm

Realistic tight-binding model for monolayer transition metal dichalcogenides in 1T' structure [article]

Mengli Hu, Guofu Ma, Chun Yu Wan, Junwei Liu
2021 arXiv   pre-print
Monolayer transition metal dichalcogenides MX_2 (M = Mo,W and X = Te, Se, S) in 1T' structure were predicted to be quantum spin Hall insulators based on first-principles calculations, which were quickly confirmed by multiple experimental groups. For a better understanding of their properties, in particular their responses to external fields, we construct a realistic four-band tight-binding (TB) model by combining the symmetry analysis and first-principles calculations. Our TB model respects all
more » ... the symmetries and can accurately reproduce the band structure in a large energy window from -0.3 eV to 0.8 eV. With the inclusion of spin-orbital coupling (SOC), our TB model can characterize the nontrivial topology and the corresponding edge states. Our TB model can also capture the anisotropic strain effects on the band structure and the strain-induced metal-insulator transition. Moreover, we found that although MX_2 share the same crystal structures and have the same crystal symmetries, while the orbital composition of states around the Fermi level are qualitatively different and their lower-energy properties cannot fully described by a single k · p model. Thus, we construct two different types of k · p model for MS_2,MSe_2 and MTe_2, respectively. Benefiting from the high accuracy and simplicity, our TB and k · p models can serve as a solid and concrete starting point for future studies of transport, superconductivity, strong correlation effects and twistronics in 1T'-transition metal dichalcogenides.
arXiv:2103.15655v1 fatcat:m5txcbx6ibbczawhn2mf4nnsbu

Fingerprint Protected Password Authentication Protocol

Chao Yang, Junwei Zhang, Jingjing Guo, Yu Zheng, Li Yang, Jianfeng Ma
2019 Security and Communication Networks  
With the rapid development of industrial Internet of things (IIOT), a variety of cloud services have been deployed to store and process the big data of IIOT. The traditional password only authentication is unable to meet the needs of security situation in IIOT. Therefore, a lot of mobile phone assisted password authentication schemes have been proposed. However, in existing schemes, the secret information is required to be stored in the user's mobile phone. Once the phone is lost, the secret
more » ... ormation may be obtained by the opponent, which will bring irreparable loss to the user. To address the above problems, we propose a fingerprint protected password authentication scheme which has no need to store the secret parameter in the mobile phone. When a user logs in, he uses his mobile phone to generate the private key which is used to decrypt the encrypted text generated during the registration phase. The process of generating the private key needs to enter the password and the fingerprint. When the computer interacts with the mobile phone, the user's password will be blinded so that it can protect the user's password from adversary's attacks. Theoretical analysis and experimental results show that our scheme improves the security of the user's secret. Meanwhile, our scheme can resist the opponent's dictionary attacks, replay attacks, and phishing attack. Our scheme can reduce the storage pressure of the mobile phone and is easy to deploy.
doi:10.1155/2019/1694702 fatcat:32uys6gnubgx7jb6yzmacqw7by

Universally composable one-time signature and broadcast authentication

JunWei Zhang, JianFeng Ma, SangJae Moon
2010 Science China Information Sciences  
Broadcast authentication is a vital security primitive for the management of a copious number of parties. In the universally composable framework, this paper investigates broadcast authentication using one-time signature based on the fact that one-time signature has efficient signature generation and verification suitable for low-power devices, and gives immediate authentication, which is a favorable property for time-critical messages. This paper first formulates a broadcast authentication
more » ... l with the ideal functionalities such as one-time signature and broadcast authentication, and proposes a broadcast authentication scheme in the hybrid model. This paper then improves HORS, which is secure based on a strong assumption (i.e., a subset-resilient hash function) and presents the improved version as HORS+, which differs from HORS such that it is a secure one-time signature based on weaker assumptions, i.e. one-way functions, one-way hash functions and collisionresistant hash functions. At the same time, a protocol OWC using one-way chains is proposed to provide more registered keys for multi-message broadcast authentication. Our broadcast authentication scheme constructed by the combined use of HORS+ and OWC is universally composable secure and suitable for low-power devices.
doi:10.1007/s11432-010-0056-2 fatcat:dp2epgjfjjcqvcp7mfsapxjwgy

Shared Biological Pathways Between Alzheimer's Disease and Ischemic Stroke

Pan Cui, Xiaofeng Ma, He Li, Wenjing Lang, Junwei Hao
2018 Frontiers in Neuroscience  
Alzheimer's disease (AD) and ischemic stroke (IS) are an immense socioeconomic burden worldwide. There is a possibility that shared genetic factors lead to their links at epidemiological and pathophysiological levels. Although recent genome-wide association studies (GWAS) have provided profound insights into the genetics of AD and IS, no shared genetic variants have been identified to date. This prompted us to initiate this study, which sought to identify shared pathways linking AD and IS. We
more » ... ok advantage of large-scale GWAS summary data of AD (17,008 AD cases and 37,154 controls) and IS (10,307 cases and 19,326 controls) to conduct pathway analyses using genetic pathways from multiple well-studied databases, including GO, KEGG, PANTHER, Reactome, and Wikipathways. Collectively, we discovered that AD and IS shared 179 GO categories (56 biological processes, 95 cellular components, and 28 molecular functions); and the following pathways: six KEGG pathways; two PANTHER pathways; four Reactome pathways; and one in Wikipathways pathway. The more fine-grained GO terms were mainly summarized into different functional categories: transcriptional and post-transcriptional regulation, synapse, endocytic membrane traffic through the endosomal system, signaling transduction, immune process, multiorganism process, protein catabolic metabolism, and cell adhesion. The shared pathways were roughly classified into three categories: immune system; cancer (NSCLC and glioma); and signal transduction pathways involving the cadherin signaling pathway, Wnt signaling pathway, G-protein signaling and downstream signaling mediated by phosphoinositides (PIPs). The majority of these common pathways linked to both AD and IS were supported by convincing evidence from the literature. In conclusion, our findings contribute to a better understanding of common biological mechanisms underlying AD and IS and serve as a guide to direct future research.
doi:10.3389/fnins.2018.00605 pmid:30245614 pmcid:PMC6137293 fatcat:s6wtkmski5fktg5mxmksfdz4yq

Overexpression of GRK3, Promoting Tumor Proliferation, Is Predictive of Poor Prognosis in Colon Cancer

Tao Jiang, Chun Yang, Liyuan Ma, Zehua Wu, Ling Ye, Xiaoqiang Ma, Hai Li, Junwei Fan, Yinxue Yang
2017 Disease Markers  
Each 50 μg aliquot of total protein was separated in 8% SDS-PAGE gel and then transferred onto polyvinylidene difluoride membranes (Millipore, Billerica, MA).  ...  Authors' Contributions Tao Jiang, Chun Yang, and Liyuan Ma contributed equally to this work. Figure 2 : 2 GRK3 expression and survival in colon cancer patients.  ... 
doi:10.1155/2017/1202710 pmid:29445249 pmcid:PMC5763208 fatcat:rckatkys4zgc5b3zrfsfe6b4vi

Sustainable Competitive Position of Mobile Communication Companies: Comprehensive Perspectives of Insiders and Outsiders

Junwei Ma, Jianhua Wang, Philip Szmedra
2019 Sustainability  
The identification of a sustainable competitive position in enterprises is a common concern of information science and sustainable development research. Achievement of sustainable competitive position and superior performance is the first priority of business organizations. However, the existing research focuses on technology competition and light market competition. The research proposes a comprehensive framework for identifying an enterprise's sustainable competitive position based on a
more » ... mensional perspective of insiders (enterprise self-positioning) and outsiders (public cognition), including enterprise technology competition identification based on technology proximity, and enterprise market competition identification based on URL co-occurrence analysis. Based on the patents collected using the Derwent Innovation Index (DII) and the URL data collected from the Internet, the sustainable competitive position of mobile communication companies is identified from two dimensions of insiders and outsiders. QAP (quadratic assignment procedure) correlation analysis shows that these two dimensions are complementary. The two-dimensional perspective is integrated to comprehensively analyze the sustainable competitive position of Huawei, and thus obtain valuable information. In order to achieve the sustainability of competitive position in the industry, mobile communication enterprises must improve their management level (marketing characteristics of management and low cost structure), and must also have the ability to break through the governance system (rigid characteristics of governance) in a timely manner, and gain the ability to create value for customers in the long run (enhancement of public cognition and public image, and customer loyalty). Meanwhile, the mobile communication industry is a typical network externality market. If the companies want to obtain a sustainable competitive position in the competitive landscape, they need to manage user's expectations, expand the foundation of users, achieve extensive cooperation and alliances with manufacturers of complementary products, and improve the quality and reputation of their brands.
doi:10.3390/su11071981 fatcat:t4iffwzctfhf5ha2d4abmyil4e
« Previous Showing results 1 — 15 out of 977 results