检索结果-南通市图书馆

Relative difficulty distillation for semantic segmentation

维普期刊数据库评论

在线全文

维普期刊数据库

学校读者我要写书评

暂无评论

Science China(Information Sciences) 2024年第9期67卷 126-145页

作者： Dong LIANG Yue SUN Yun DU Songcan CHEN sheng-jun huang MIIT Key Laboratory of Pattern Analysis and Machine Intelligence College of Computer Science and Technology Nanjing University of Aeronautics and Astronautics Shenzhen Research Institute Nanjing University of Aeronautics and Astronautics

Current knowledge distillation(KD) methods primarily focus on transferring various structured knowledge and designing corresponding optimization goals to encourage the student network to imitate the output of the teacher network. However, introducing too many additional optimization objectives may lead to unstable training, such as gradient conflicts. Moreover, these methods ignored the guidelines of relative learning difficulty between the teacher and student networks. Inspired by human cognitive science, in this paper, we redefine knowledge from a new perspective — the student and teacher networks' relative difficulty of samples, and propose a pixel-level KD paradigm for semantic segmentation named relative difficulty distillation(RDD). We propose a two-stage RDD framework: teacher-full evaluated RDD(TFE-RDD) and teacher-student evaluated RDD(TSE-RDD). RDD allows the teacher network to provide effective guidance on learning focus without additional optimization goals, thus avoiding adjusting learning weights for multiple losses. Extensive experimental evaluations using a general distillation loss function on popular datasets such as Cityscapes, Cam Vid, Pascal VOC, and ADE20k demonstrate the effectiveness of RDD against state-ofthe-art KD methods. Additionally, our research showcases that RDD can integrate with existing KD methods to improve their upper performance bound. Codes are available at https://***/sunyueue/***.

关键词： knowledge distillation semantic segmentation relative difficulty sample weighting prediction discrepancy

Crowdsourcing aggregation with deep Bayesian learning

在线全文

同方期刊数据库

学校读者我要写书评

暂无评论

Science China(Information Sciences) 2021年第3期64卷 46-56页

作者： Shao-Yuan LI sheng-jun huang Songcan CHEN College of Computer Science and Technology College of Artificial IntelligenceNanjing University of Aeronautics and Astronautics State Key Laboratory for Novel Software Technology Nanjing University

In this study, we consider a crowdsourcing classification problem in which labeling information from crowds is aggregated to infer latent true labels. We propose a fully Bayesian deep generative crowdsourcing model(Bayes DGC), which combines the strength of deep neural networks(DNNs) on automatic representation learning and the interpretable probabilistic structure encoding of probabilistic graphical models. The model comprises a DNN classifier as a prior for the true labels and a probabilistic model for the annotation generation process. The DNN classifier and annotation generation process share the latent true label variables. To address the inference challenge, we developed a natural-gradient stochastic variational inference, which combines variational message passing for conjugate parameters and stochastic gradient descent for DNN and learns the distribution of latent true labels and workers’ confusion matrix via end-to-end training. We illustrated the effectiveness of the proposed model using empirical results on 22 real-world datasets.

关键词： crowdsourcing classification fully Bayesian deep generative models natural gradient stochastic variational inference

Multi-label active learning by model guided distribution matching

在线全文

同方期刊数据库

学校读者我要写书评

暂无评论

Frontiers of Computer Science 2016年第5期10卷 845-855页

作者： Nengneng GAO sheng-jun huang Songcan CHEN College of Computer Science and Technology Nanjing University of Aeronautics and AstronauticsCollaborative Innovation Center of Novel Software Technology and Industrialization Nanjing 211106 China

Multi-label learning is an effective framework for learning with objects that have multiple semantic labels, and has been successfully applied into many real-world tasks, In contrast with traditional single-label learning, the cost of la- beling a multi-label example is rather high, thus it becomes an important task to train an effective multi-label learning model with as few labeled examples as possible. Active learning, which actively selects the most valuable data to query their labels, is the most important approach to reduce labeling cost. In this paper, we propose a novel approach MADM for batch mode multi-label active learning. On one hand, MADM exploits representativeness and diversity in both the feature and label space by matching the distribution between labeled and unlabeled data. On the other hand, it tends to query predicted positive instances, which are expected to be more informative than negative ones. Experiments on benchmark datasets demonstrate that the proposed approach can reduce the labeling cost significantly.

关键词： multi-label learning batch mode active learning distribution matching

BL03HB: a Laue microdiffraction beamline for both protein crystallography and materials science at SSRF

在线全文

学校读者我要写书评

暂无评论

Nuclear Science and Techniques 2024年第7期35卷 68-78页

作者： Zhi-jun Wang Si-sheng Wang Zheng-huang Su Li Yu Yu-Zhu Wang Bo Sun Wen Wen Xing-Yu Gao Shanghai Advanced Research Institute Chinese Academy of SciencesShanghai201210China University of Chinese Academy of Science Beijing100049China

A Laue microdiffraction beamline(BL03HB) was constructed at the Shanghai Synchrotron Radiation Facility(SSRF).This beamline features two consecutive focusing points in two different sectors within its end station, the first dedicated to protein crystallography and the other tailored to materials science applications. Based on a superbend dipole magnet with a magnetic field of 2.29 T, a two-stage focusing design was implemented with two sets of Kirkpatrick-Baez mirrors to achieve a micro white beam as small as 4.2 μ m ×4.3 μ m at the first sector and 0.9 μ m ×1.3 μ m at the second sector in the standard beamline operation mode at SSRF. The X-ray microbeam in the two sectors can be easily switched between monochromatic and white beams by moving a four-bounce monochromator in or out of the light path, respectively. In the protein crystallography sector, white-beam Laue microdiffraction was demonstrated to successfully determine the structure of protein crystals from only a few images of diffraction data collected by a Pilatus 2 M area detector. In the materials science sector,the white-beam Laue diffraction was collected in a reflection geometry using another Pilatus 2 M area detector, which could map the microstructural distribution on the sample surface by scanning the samples. In general, the BL03HB beamline promotes the application of Laue microdiffraction in both protein crystallography and materials science. This paper presents a comprehensive overview of the BL03HB beamline, end station, and the first commission results.

关键词： Crystallography White-beam Laue microdiffraction Beamline Synchrotron

Chimeric Monkey Born Alive with a High Contribution of Donor Cells

在线全文

学校读者我要写书评

暂无评论

Neuroscience Bulletin 2024年第6期40卷 849-851页

作者： Jia huang jun Wu Department of Molecular Biology University of Texas Southwestern Medical CenterDallasTX75390USA Hamon Center for Regenerative Science and Medicine University of Texas Southwestern Medical CenterDallasTX75390USA Cecil H.and Ida Green Center for Reproductive Biology Sciences University of Texas Southwestern Medical CenterDallasTX75390USA

Chimeras,which consist of cells from different individuals or different species,represent an important tool in biological *** in the early twentieth century,studies on chimeras have been focusing on understanding fundamental genetic and cellular developmental *** the advances in genetic engineering and embryonic stem cell(ESC)technology,chimeras have become an important tool for the generation of transgenic animals.

关键词： mechanisms. advances developmental

Energy-information trade-off induces continuous and discontinuous phase transitions in lateral predictive coding

维普期刊数据库评论

在线全文

维普期刊数据库

学校读者我要写书评

暂无评论

Science China(Physics,Mechanics & Astronomy) 2024年第6期67卷 79-85页

作者： Zhen-Ye huang Ruyi Zhou Miao huang Hai-jun Zhou Key Laboratory for Theoretical Physics Institute of Theoretical PhysicsChinese Academy of SciencesBeijing 100190China School of Physical Sciences University of Chinese Academy of SciencesBeijing 100049China School of Optics and Photonics Beijing Institute of TechnologyBeijing 100081China Minjiang Collaborative Center for Theoretical Physics Minjiang UniversityFuzhou 350108China

Lateral predictive coding is a recurrent neural network that creates energy-efficient internal representations by exploiting statistical regularity in sensory ***,we analytically investigate the trade-off between information robustness and energy in a linear model of lateral predictive coding and numerically minimize a free energy *** observed several phase transitions in the synaptic weight matrix,particularly a continuous transition that breaks reciprocity and permutation symmetry and builds cyclic dominance and a discontinuous transition with the associated sudden emergence of tight balance between excitatory and inhibitory *** optimal network follows an ideal gas law over an extended temperature range and saturates the efficiency upper bound of energy *** results provide theoretical insights into the emergence and evolution of complex internal models in predictive processing systems.

关键词： predictive coding recurrent neural network phase transition internal model free energy

同方期刊数据库

Revealing the Structure-Property Difference of Natural Rubber Prepared by Different Methods: Protein and Gel Content are Key Factors

在线全文

学校读者我要写书评

暂无评论

Chinese Journal of Polymer Science 2024年第4期42卷 457-467页

作者： Shao-Qi huang jun-Qi Zhang Yong Zhu Ling-Min Kong Lu-sheng Liao Fu-Quan Zhang Zheng-Tian Xie Jin-Rong Wu State Key Laboratory of Polymer Materials Engineering College of Polymer Science and EngineeringSichuan UniversityChengdu610065China Hainan Provincial Key Laboratory of Natural Rubber Processing Agricultural Products Processing Research Institute of Chinese Academy of Tropical Agricultural SciencesZhanjiang524001China

Natural rubber (NR) is widely used in various fields including aerospace, military industry and transportation due to its superior elasticity and comprehensive mechanical properties. Nonetheless, the commercial NR prepared by different methods usually exhibits different mechanical properties, primarily due to variations in processing conditions during the conversion from latex to bulk rubber material. Consequently, this poses challenges in scientific research and industrial production of NR. In order to assess the properties of various commercially available NR and identify key structural and compositional components, this study systematically compares and analyzes four representative NR raw materials: air dried sheet (ADS), ribbed smoked sheets (RSS), constant viscidity NR (CV), and whole field latex rubber (WF). The investigation focuses on evaluating their static mechanical behavior, SIC behavior, wear resistance, and fatigue resistance. The findings indicate that protein and gel content exhibit a crucial influence on the NR properties. These constituents contribute to the formation of a high-crosslinking density region, generating a heterogeneous network structure within the rubber. This structure amplifies strains during deformation, leading to earlier and stronger strain-induced crystallization (SIC). Among the four commercial NR brands, RSS demonstrates superior overall mechanical and dynamic properties owing to its high protein and gel content. This study serves as a valuable reference for comprehending the differences in properties among various commercial NR, thereby offering guidance for the actual processing and selection of NR.

关键词： Natural rubber Processingmethod Protein Gel content Property

The occurrence characteristics of oil in shales matrix from organic geochemical screening data and pore structure properties:An experimental study Author links open overlay panel

在线全文

学校读者我要写书评

暂无评论

Petroleum Science 2024年第1期21卷 1-13页

作者： Zi-Zhi Lin jun-Qian Li Shuang-Fang Lu Qin-Hong Hu Peng-Fei Zhang jun-Jie Wang Qi Zhi Hong-sheng huang Na Yin Yue Wang Tian-Chen Ge National Key Laboratory of Deep Oil and Gas China University of Petroleum(East China)Qingdao 266580ShandongChina School of Geosciences China University of Petroleum(East China)Qingdao 266580ShandongChina Sanya Offshore Oil&Gas Research Institute Northeast Petroleum UniversitySanya 572025HainanChina College of Earth Science and Engineering Shandong University of Science and TechnologyQingdao 266590ShandongChina

The occurrence characteristics of shale oil are of great significance to the movability of shale *** this study,the occurrence characteristics of oil in the shale matrix at Funing Formation shale in Subei Basin were quantitatively evaluated by organic geochemistry and microscopic pore structure characterization *** Multiple Isothermal Stages Pyrolysis(MIS)experiment results show that the content of total oil,adsorbed oil,and free oil in the shales are 3.15-11.25 mg/g,1.41-4.95 mg/g,and 1.74-6.51 mg/g,*** which the silicon-rich shale has the best *** relative content of free oil shows an increasing trend in pores with pore diameters greater than 3 *** the relative content of free oil reaches 100%,the pore size of silicon-rich shale is about 200 nm,while that of calcium-rich shale,clay-rich shale,and siliceous mixed shale is about 10 *** occurrence law of adsorbed oil is opposite to that of free oil,which indicates that shale oil will occur in the pores and fractures in a free state in a more extensive pore size range(>200 nm).This study also enables us to further understand the occurrence characteristics of shale oil under the interaction of occurrence state and occurrence space.

关键词： Shale oil Occurrence characteristics Sequential solvent extraction Low-temperature nitrogen adsorption Funing Formation Dongtai Depression

Prognostic significance and relationship of SMAD3 phosphoisoforms and VEGFR-1 in gastric cancer:A clinicopathological study

在线全文

学校读者我要写书评

暂无评论

World Journal of Gastrointestinal Oncology 2024年第1期16卷 118-132页

作者： Shi-Lin Lv Pei Guo jun-Rong Zou Ren-sheng Chen Ling-Yu Luo De-Qiang huang Hospital of Gastroenterology Institute of Digestive DiseasesThe First Affiliated Hospital of Nanchang UniversityNanchang 330006Jiangxi ProvinceChina Department of Pathology Shenzhen Hospital of Southern Medical UniversityShenzhen 518000Guangdong ProvinceChina Institute of Urology The First Affiliated Hospital of Gannan Medical UniversityGanzhou 341000Jiangxi ProvinceChina Department of Pathology The First Affiliated Hospital of Nanchang UniversityNanchang 330006Jiangxi ProvinceChina

BACKGROUND The TGF-β/SMAD3 and VEGFR-1 signaling pathways play important roles in gastric cancer ***3 phosphorylation is a crucial prognostic marker in gastric *** To determine the prognostic value and relationship of SMAD3 phospho-isoforms and VEGFR-1 in gastric *** This was a single-center observational study which enrolled 98 gastric cancer patients and 82 adjacent normal gastric tissues from patients aged 32-84 years(median age 65)between July 2006 and April *** were followed up until death or the study ended(median follow-up duration of 28.5 mo).The samples were used to generate tissue microarrays(TMAs)for immunohistochemical(IHC)*** expressions of TGF-β1,pSMAD3C(S423/425),pSMAD3L(S204),and VEGFR-1 in gastric cancer(GC)tumor tissue and normal tissue were measured by IHC staining using TMAs obtained from 98 GC *** and survival information of the patients was recorded by Outdo Biotech from May 2007 to July *** relationship between TGF-β1,pSMAD3C(S423/425),pSMAD3L(S204),and VEGFR-1 protein expression levels was analyzed using Pearson's correlation *** relationship between protein expression levels and clinicopathological parameters was analyzed using the Chi-squared test.A survival curve was generated using the Kaplan-Meier survival *** TGFβ-1 and VEGFR-1 expression was significantly upregulated in gastric cancer tissue compared to adjacent noncancerous *** positive expression of phosphorylated isoforms of Smad3 varied depending on the phosphorylation site[pSMAD3C(S423/425):51.0%and pSMAD3L(S204):31.6%].High expression of pSMAD-3L(S204)was significantly correlated with larger tumors(P=0.038)and later N stages(P=0.035).Additionally,high expression of VEGFR-1 was closely correlated with tumor size(P=0.015)and pathological grading(P=0.013).High expression of both pSMAD3L(S204)and VEGFR-1 was associated with unfavorable outcomes in terms of overall survival(OS).Multiva

关键词： Gastric cancer pSMAD3L(S204) pSMAD3C(S423/425) Survival Transforming growth factor-β1 VEGFR-1