Search

In person re-identification (re-ID) task, it is still challenging to learn discriminative representation by deep learning, due to limited data. Generally speaking, the model will get better performance when increasing the amount of data. The addition of similar classes strengthens the ability of the classifier to identify similar identities, thereby improving the discrimination of representation. In this paper, we propose a Diverse and Compact Transformer (DC-Former) that can achieve a similar effect by splitting embedding space into multiple diverse and compact subspaces. Compact embedding subspace helps model learn more robust and discriminative embedding to identify similar classes. And the fusion of these diverse embeddings containing more fine-grained information can further improve the effect of re-ID. Specifically, multiple class tokens are used in vision transformer to represent multiple embedding spaces. Then, a self-diverse constraint (SDC) is applied to these spaces to push them away from each other, which makes each embedding space diverse and compact. Further, a dynamic weight controller(DWC) is further designed for balancing the relative importance among them during training. The experimental results of our method are promising, which surpass previous state-of-the-art methods on several commonly used person re-ID benchmarks., Comment: Accepted by AAAI23

Embedding models have shown great power in knowledge graph completion (KGC) task. By learning structural constraints for each training triple, these methods implicitly memorize intrinsic relation rules to infer missing links. However, this paper points out that the multi-hop relation rules are hard to be reliably memorized due to the inherent deficiencies of such implicit memorization strategy, making embedding models underperform in predicting links between distant entity pairs. To alleviate this problem, we present Vertical Learning Paradigm (VLP), which extends embedding models by allowing to explicitly copy target information from related factual triples for more accurate prediction. Rather than solely relying on the implicit memory, VLP directly provides additional cues to improve the generalization ability of embedding models, especially making the distant link prediction significantly easier. Moreover, we also propose a novel relative distance based negative sampling technique (ReD) for more effective optimization. Experiments demonstrate the validity and generality of our proposals on two standard benchmarks. Our code is available at https://github.com/rui9812/VLP., Comment: Accepted to ACL 2023 Main Conference (Long Paper)

This paper studies learning on text-attributed graphs (TAGs), where each node is associated with a text description. An ideal solution for such a problem would be integrating both the text and graph structure information with large language models and graph neural networks (GNNs). However, the problem becomes very challenging when graphs are large due to the high computational complexity brought by training large language models and GNNs together. In this paper, we propose an efficient and effective solution to learning on large text-attributed graphs by fusing graph structure and language learning with a variational Expectation-Maximization (EM) framework, called GLEM. Instead of simultaneously training large language models and GNNs on big graphs, GLEM proposes to alternatively update the two modules in the E-step and M-step. Such a procedure allows training the two modules separately while simultaneously allowing the two modules to interact and mutually enhance each other. Extensive experiments on multiple data sets demonstrate the efficiency and effectiveness of the proposed approach., ICLR 2023

Graph Neural Networks (GNNs) have made tremendous progress in the graph classification task. However, a performance gap between the training set and the test set has often been noticed. To bridge such gap, in this work we introduce the first test-time training framework for GNNs to enhance the model generalization capacity for the graph classification task. In particular, we design a novel test-time training strategy with self-supervised learning to adjust the GNN model for each test graph sample. Experiments on the benchmark datasets have demonstrated the effectiveness of the proposed framework, especially when there are distribution shifts between training set and test set. We have also conducted exploratory studies and theoretical analysis to gain deeper understandings on the rationality of the design of the proposed graph test time training framework (GT3).

ObjectiveTo investigate the effect of Cordyceps polysaccharides on hepatocyte apoptosis and its mechanism of action. Methods Normal human L02 hepatocytes without any drug treatment was established as normal group. The L02 hepatocytes were treated with different concentrations of tumor necrosis factor-α (TNF-α) (5, 10, 20, and 40 ng/ml) for 24 hours to screen out the modeling conditions for the model of hepatocyte apoptosis, which were established as model group. According to the experimental design, L02 hepatocytes pretreated with three different concentrations of Cordyceps polysaccharides (50, 100, and 200 μg/ml) for 12 hours, with or without TNF-α treatment for 24 hours at the selected concentration, were established as experimental group. Related samples were collected for analysis. CCK8 assay was used to measure cell proliferation; Annexin V/PI double staining was used to measure the number of apoptotic hepatocytes; RT-PCR was used to measure the mRNA expression of cell apoptosis genes (Bax, caspase-3, caspase-8, caspase-9, and death receptor Fas); Western blotting was used to measure the protein expression of cleaved-caspase-3 and cleaved-caspase-8. A one-way analysis of variance or the Kruskal-Wallis H test was used for comparison of continuous data between multiple groups, and the least significant difference t-test or the Dunnett-t test was used for further comparison between two groups. ResultsCCK8 assay showed that compared with the normal group, the L02 hepatocytes induced by 40 ng/ml TNF-α for 24 hours had a significant reduction in proliferation (73.54%±14.19% vs 100.00%±23.61%, P＜0.01), and compared with the model group under the premise that Cordyceps polysaccharides had no obvious cytotoxicity to hepatocytes at the three concentrations, the 50, 100, and 200 μg/ml experimental groups had a significant increase in cell proliferation (108.10%±9.05%/114.30%±8.79%/117.70%±9.66% vs 93.02%±7.21%, all P＜0.01). Annexin V/PI double staining showed that compared with the normal group, the model group had a significant increase in the number of apoptotic cells (11.57%±1.41% vs 7.71%±1.20%, P＜0.05), and compared with the model group, the 50, 100, and 200 μg/ml experimental groups had a significant reduction in the number of apoptotic cells (1516%±0.16%/13.28%±1.57%/16.91%±0.21% vs 18.91%±0.80%, all P＜0.05). RT-PCR and Western blotting showed that compared with the normal group, the L02 hepatocytes induced by 40 ng/ml TNF-α had significant increases in the mRNA expression of the apoptosis-related genes Bax, caspase-9, and death receptor Fas (all P＜005), as well as significant increases in the protein expression of the activated forms of cleaved-caspase-3 and cleaved-caspase-8 (P＜0.05). Compared with the model group, the 50 μg/ml experimental group had significant reductions in the mRNA expression of Bax, caspase-3, and caspase-9 and the protein expression of cleaved-caspase-8; the 100 μg/ml experimental group had significant reductions in the mRNA expression of Bax, caspase-3, caspase-8, Fas, and caspase-9 and the protein expression of cleaved-caspase-8; the 200 μg/ml experimental group had significant reductions in the mRNA expression of caspase-3, caspase-8, Fas, and caspase-9 and the protein expression of cleaved-caspase-3 and cleaved-caspase-8. The three concentrations of Cordyceps polysaccharides had a certain dose-effect trend. Conclusion Cordyceps polysaccharides can effectively inhibit the apoptosis of normal L02 hepatocytes induced by TNF-α.

The effectiveness of knowledge graph embedding (KGE) largely depends on the ability to model intrinsic relation patterns and mapping properties. However, existing approaches can only capture some of them with insufficient modeling capacity. In this work, we propose a more powerful KGE framework named HousE, which involves a novel parameterization based on two kinds of Householder transformations: (1) Householder rotations to achieve superior capacity of modeling relation patterns; (2) Householder projections to handle sophisticated relation mapping properties. Theoretically, HousE is capable of modeling crucial relation patterns and mapping properties simultaneously. Besides, HousE is a generalization of existing rotation-based models while extending the rotations to high-dimensional spaces. Empirically, HousE achieves new state-of-the-art performance on five benchmark datasets. Our code is available at https://github.com/anrep/HousE., Comment: Accepted by ICML 2022

Triplet loss is a widely adopted loss function in ReID task which pulls the hardest positive pairs close and pushes the hardest negative pairs far away. However, the selected samples are not the hardest globally, but the hardest only in a mini-batch, which will affect the performance. In this report, a hard batch mining method is proposed to mine the hardest samples globally to make triplet harder. More specifically, the most similar classes are selected into a same mini-batch so that the similar classes could be pushed further away. Besides, an adversarial scene removal module composed of a scene classifier and an adversarial loss is used to learn scene invariant feature representations. Experiments are conducted on dataset MSMT17 to prove the effectiveness, and our method surpasses all of the previous methods and sets state-of-the-art result.

Transformers have achieved remarkable performance in a myriad of fields including natural language processing and computer vision. However, when it comes to the graph mining area, where graph neural network (GNN) has been the dominant paradigm, transformers haven't achieved competitive performance, especially on the node classification task. Existing graph transformer models typically adopt fully-connected attention mechanism on the whole input graph and thus suffer from severe scalability issues and are intractable to train in data insufficient cases. To alleviate these issues, we propose a novel Gophormer model which applies transformers on ego-graphs instead of full-graphs. Specifically, Node2Seq module is proposed to sample ego-graphs as the input of transformers, which alleviates the challenge of scalability and serves as an effective data augmentation technique to boost model performance. Moreover, different from the feature-based attention strategy in vanilla transformers, we propose a proximity-enhanced attention mechanism to capture the fine-grained structural bias. In order to handle the uncertainty introduced by the ego-graph sampling, we further propose a consistency regularization and a multi-sample inference strategy for stabilized training and testing, respectively. Extensive experiments on six benchmark datasets are conducted to demonstrate the superiority of Gophormer over existing graph transformers and popular GNNs, revealing the promising future of graph transformers.

Intelligent nursing crib is a nursing product that provides different services to infants of different ages on the platform of crib. It uses generic cabling technology, behavioral intelligent analysis technology, temperature, humidity, audio discrimination technology, automatic warning technology, network communication technology and crib protection facilities. To construct the early warning system of observing the infant's state and monitoring the infant's dangerous state, reduce the risk of infant accidents, and reduce the pressure of parents on nursing.

In machine learning, it is specially important to determine those variables that are relevant to the object of study.In particular, the data sets usually used in software engineering have a high number of variables and researchers andpractitioners should select those that are most relevant as independent variables for effort estimation purposes.The project seeks to know how these algorithms are implemented, particularly those based on information theory ofShannon (entropy, mutual information and redundancy). The implementation language will be R. From this implementation,the designer will adapt some of these algorithms to improve their performance., En machine learning es especialmente importante determinas las variables que son relevantes al objeto de estudio. En particular, los data sets utilizados habitualmente en ingeniería del software tienen un alto número de variables y tanto investigadores como profesionales deberían seleccionar aquellos que son más relevantes como variables independientes para propósitos de estimación de esfuerzo. El proyecto pretende conocer como están implantados estos algoritmos, fundamentalmente aquellos que están basados en en la teoría de información de Shannon (entropía, información mutua y redundancia). El lenguaje de implementación será R. A partir de esta implementación, el proyectista adaptará algunos de estos algoritmos para mejorar su desempeño.

To devise low-carbon cement clinker product, in-depth knowledge about the variation mechanism of cement clinker CO2 emissions is the prerequisite. The determinants of cement clinker CO2 emissions in Chinese cement industry have been inquired by several researches recently. However, the impacts and abatement effect of clinker quality on cement clinker CO2 emissions have not been identified yet. In technological sense, there lacks an integrated quantitative model that can elucidate the dynamic mechanism between cement clinker CO2 emissions and clinker quality. To fill this gap, this paper devises a CO2 emission accounting framework integrated with optimization model to explore the dynamic mechanism between cement clinker CO2 emissions and clinker quality. Based on survey data of 2 typical cement clinker production lines in China, the accounting framework is validated with measured data. Through numerical simulation within this framework, a set of sensitivity analysis is conducted. The results indicate that if the target KH (lime saturation ratio) of both production lines climb from -5% to +5%, the process CO2 emission factor will give rise to an increase of about 16.4 and 14.8 kg CO2/t clinker, respectively. With the same variation, the fuel CO2 emission factors of two production lines will gain 17.2 and 13.7 kg CO2/t clinker, respectively. Obviously, controlled decrease of clinker quality can reduce CO2 emission of clinker manufacturing. The results of sensitivity analysis are robust and highlight the remarkable role of clinker quality in both process and fuel emissions. Besides, the integrated framework can provide in-depth investigation into the dynamic mechanism between CO2 emissions and the other determinants. Using this framework, cement plants or industry can develop optimal clinker quality standards on a quantitative basis to achieve low-carbon clinker. (C) 2016 Elsevier B.V. All rights reserved.

CO2 emission estimation of cement production is an underlying tool to identify the CO2 sources, to evaluate the reduction efforts and to set prospective reduction targets for the Chinese cement industry. However, lack of domestic data has limited the reliability and accuracy of CO2 emission estimation for Chinese cement industry. To develop an accurate and comprehensive CO2 emission factor for Chinese cement industry, this study established a factory-level database of 197 cement production lines from 21 provinces covering various capacity scales. On the basis of this database, process, fuel, electricity and synthesized emission factor were computed. Furthermore, bootstrap simulation and Monte Carlo simulation were applied to evaluate the uncertainty of these factors. After corrections for cement kiln dust (CKD), incomplete decomposition, organic carbon and inorganic carbon, the medians of process, fuel and direct emission factors are 525, 369, and 919 kg CO2/t clinker, respectively. Electricity emission factor is 74.9 kg CO2/t clinker. The final synthesized emission factor for cement product is 761 kg CO2/t cement with uncertainties of [-34.8%, +31.69]. In this study, two revised calculation methods for the process emission factor are applied. Two calculation methods for the fuel emission factor are adopted as well. These practices seek to improve the reliability and accuracy of cement CO2 emission factor. The simulated results indicate that the revised output method produces more accurate estimation for the process emission factor than the revised input method and unrevised output method. For fuel emission factor, the CC (carbon content) method is more accurate than the NCV (net calorific value) method. Simulated results are also compared with other authoritative estimation to validate the reliability of the CO2 emission factors calculated in the present study. (C) 2016 Elsevier Ltd. All rights reserved.

The thesis for the Master of Music degree in Orchestral Instruments consists two full-length recitals, or the equivalent, with emphasis on solo works, but also including representative chamber works. The two full-length recitals were performed on April 9, 2005 and April 18, 2005.