Osteoarthritis (OA), the most prevalent age-related joint disorder, is characterized by chronic inflammation, progressive articular cartilage destruction, and subchondral bone sclerosis. Accumulating ...evidences indicate that circular RNAs (circRNAs) play a critical role in various diseases, but the function of circRNAs in OA remains largely unknown. Here we showed that circRNA.33186 was significantly upregulated in IL-1β)-treated chondrocytes and in cartilage tissues of a destabilized medial meniscus (DMM)-induced OA mouse model. Knockdown of circRNA.33186 increased anabolic factor (type II collagen) expression and decreased catabolic factor (MMP-13) expression. Knockdown of circRNA.33186 also promoted proliferation and inhibited apoptosis in IL-1β-treated chondrocytes. Silencing of circRNA.33186 in vivo markedly alleviated DMM-induced OA. Mechanistic study showed that circRNA.33186 directly binds to and inhibits miR-127-5p, thereby increasing MMP-13 expression, and contributes to OA pathogenesis. Taken together, our findings demonstrated a fundamental role of circRNA.33186 in OA progression and provide a potential drug target in OA therapy.
Zhu and colleagues demonstrate that circRNA.33186 regulates chondrocyte functions, including ECM catabolism, proliferation, and apoptosis. Silencing of circRNA.33186 alleviated OA by acting as a sponge of miR-127-5p. These findings reveal a fundamental role of circRNA.33186 in OA progression and provide a potential drug target in OA therapy.
Domain adaptation manages to build an effective target classifier or regression model for unlabeled target data by utilizing the well-labeled source data but lying different distributions. ...Intuitively, to address domain shift problem, it is crucial to learn domain invariant features across domains, and most existing approaches have concentrated on it. However, they often do not directly constrain the learned features to be class discriminative for both source and target data, which is of vital importance for the final classification. Therefore, in this paper, we put forward a novel feature learning method for domain adaptation to construct both domain invariant and class discriminative representations, referred to as DICD. Specifically, DICD is to learn a latent feature space with important data properties preserved, which reduces the domain difference by jointly matching the marginal and class-conditional distributions of both domains, and simultaneously maximizes the inter-class dispersion and minimizes the intra-class scatter as much as possible. Experiments in this paper have demonstrated that the class discriminative properties will dramatically alleviate the cross-domain distribution inconsistency, which further boosts the classification performance. Moreover, we show that exploring both domain invariance and class discriminativeness of the learned representations can be integrated into one optimization framework, and the optimal solution can be derived effectively by solving a generalized eigen-decomposition problem. Comprehensive experiments on several visual cross-domain classification tasks verify that DICD can outperform the competitors significantly.
The effects of chordwise deformation and the half-amplitude asymmetry on the hydrodynamic performance and vortex dynamics of batoid fish have been numerically investigated, in which the two ...parameters were represented by the wavenumber ($W$) and the ratio of the half-amplitude above the longitudinal axis to that below ($HAR$). Fin kinematics were prescribed based on biological data. Simulations were conducted using the immersed boundary method. It was found that moderate chordwise deformation enhances the thrust, saves the power and increases the efficiency. A large $HAR$ can also increase thrust performance. By using the derivative-moment transformation theory at several subdomains to capture the local vortical structures and a force decomposition, it was shown that, at high Strouhal numbers ($St$), the tip vortex is the main source of thrust, whereas the leading-edge vortex (LEV) and trailing-edge vortex weaken the thrust generation. However, at lower $St$, the LEV would enhance the thrust. The least deformation ($W=0$) leads to the largest effective angle of attack, and thus the strongest vortices. However, moderate deformation ($W=0.4$) has an optimal balance between the performance enhancement and the opposite effect of different local structures. The performance enhancement of $HAR$ was also due to the increase of the vortical contributions. This work provides a new insight into the role of vortices and the force enhancement mechanism in aquatic swimming.
The deployment of deep convolutional neural networks (CNNs) in many real world applications is largely hindered by their high computational cost. In this paper, we propose a novel learning scheme for ...CNNs to simultaneously 1) reduce the model size; 2) decrease the run-time memory footprint; and 3) lower the number of computing operations, without compromising accuracy. This is achieved by enforcing channel-level sparsity in the network in a simple but effective way. Different from many existing approaches, the proposed method directly applies to modern CNN architectures, introduces minimum overhead to the training process, and requires no special software/hardware accelerators for the resulting models. We call our approach network slimming, which takes wide and large networks as input models, but during training insignificant channels are automatically identified and pruned afterwards, yielding thin and compact models with comparable accuracy. We empirically demonstrate the effectiveness of our approach with several state-of-the-art CNN models, including VGGNet, ResNet and DenseNet, on various image classification datasets. For VGGNet, a multi-pass version of network slimming gives a 20× reduction in model size and a 5× reduction in computing operations.
Prediction Reweighting for Domain Adaptation Li, Shuang; Song, Shiji; Huang, Gao
IEEE transaction on neural networks and learning systems,
07/2017, Letnik:
28, Številka:
7
Journal Article
There are plenty of classification methods that perform well when training and testing data are drawn from the same distribution. However, in real applications, this condition may be violated, which ...causes degradation of classification accuracy. Domain adaptation is an effective approach to address this problem. In this paper, we propose a general domain adaptation framework from the perspective of prediction reweighting, from which a novel approach is derived. Different from the major domain adaptation methods, our idea is to reweight predictions of the training classifier on testing data according to their signed distance to the domain separator, which is a classifier that distinguishes training data (from source domain) and testing data (from target domain). We then propagate the labels of target instances with larger weights to ones with smaller weights by introducing a manifold regularization method. It can be proved that our reweighting scheme effectively brings the source and target domains closer to each other in an appropriate sense, such that classification in target domain becomes easier. The proposed method can be implemented efficiently by a simple two-stage algorithm, and the target classifier has a closed-form solution. The effectiveness of our approach is verified by the experiments on artificial datasets and two standard benchmarks, a visual object recognition task and a cross-domain sentiment analysis of text. Experimental results demonstrate that our method is competitive with the state-of-the-art domain adaptation algorithms.
Osteoarthritis (OA) is the most common joint disease and is mainly characterized by degradation of the articular cartilage. Recently, circular RNAs (circRNAs), novel noncoding RNAs with different ...biological functions and pathological implications, have been reported to be closely associated with various diseases. Growing evidence indicates that circRNAs act as competing endogenous RNAs (ceRNAs) that bind with microRNAs (miRNAs) and regulate their downstream functions. Here, we identified a new circRNA, circRNA_Atp9b, and further investigated its function in OA using a well-established mouse chondrocyte model. We demonstrated that circRNA_Atp9b expression was significantly up-regulated in mouse chondrocytes after stimulation with interleukin-1 beta (IL-1β), and that knockdown of circRNA_Atp9b promoted the expression of type II collagen while inhibiting the generation of MMP13, COX-2 and IL-6. Moreover, there was a negative correlation between the expression levels of circRNA_Atp9b and microRNA (miR)-138-5p, indicating that miR-138-5p also played a role in IL-1β-induced chondrocytes. Bioinformatics analysis predicted circRNA_Atp9b directly target miR-138-5p, which was validated by dual-luciferase assay. Further functional experiments revealed that down-regulation of miR-138-5p partly reversed the effects of circRNA_Atp9b on extracellular matrix (ECM) catabolism and inflammation. Taken together, these results suggest that circRNA_Atp9b regulates OA progression by modulating ECM catabolism and inflammation in chondrocytes via sponging miR-138-5p. Our findings provide novel insight into the regulatory mechanism of circRNA_Atp9b in OA and may contribute to establishing potential therapeutic strategies.
•CircRNA_Atp9b expression levels are up-regulated in IL-1β-induced chondrocytes.•CircRNA_Atp9b plays important roles in IL-1β-induced chondrocytes.•CircRNA_Atp9b functions as a sponge of miR-138-5p.•The functional relationship between miR-138-5p and circRNA_Atp9 in IL-1β-induced chondrocytes.
We report a low-cost, high-throughput benchtop method that enables thin layers of metal to be shaped with nanoscale precision by generating ultrahigh-strain-rate deformations. Laser shock imprinting ...can create three-dimensional crystalline metallic structures as small as 10 nanometers with ultrasmooth surfaces at ambient conditions. This technique enables the successful fabrications of large-area, uniform nanopatterns with aspect ratios as high as 5 for plasmonic and sensing applications, as well as mechanically strengthened nanostructures and metal-graphene hybrid nanodevices.
Domain adaptation (DA) attempts to transfer knowledge learned in the labeled source domain to the unlabeled but related target domain without requiring large amounts of target supervision. Recent ...advances in DA mainly proceed by aligning the source and target distributions. Despite the significant success, the adaptation performance still degrades accordingly when the source and target domains encounter a large distribution discrepancy. We consider this limitation may attribute to the insufficient exploration of domain-specialized features because most studies merely concentrate on domain-general feature learning in task-specific layers and integrate totally-shared convolutional networks (convnets) to generate common features for both domains. In this paper, we relax the completely-shared convnets assumption adopted by previous DA methods and propose Domain Conditioned Adaptation Network (DCAN) , which introduces domain conditioned channel attention module with a multi-path structure to separately excite channel activation for each domain. Such a partially-shared convnets module allows domain-specialized features in low-level to be explored appropriately. Further, given the knowledge transferability varying along with convolutional layers, we develop Generalized Domain Conditioned Adaptation Network (GDCAN) to automatically determine whether domain channel activations should be separately modeled in each attention module. Afterward, the critical domain-specialized knowledge could be adaptively extracted according to the domain statistic gaps. As far as we know, this is the first work to explore the domain-wise convolutional channel activations separately for deep DA networks. Additionally, to effectively match high-level feature distributions across domains, we consider deploying feature adaptation blocks after task-specific layers, which can explicitly mitigate the domain discrepancy. Extensive experiments on four cross-domain benchmarks, including DomainNet, Office-Home, Office-31, and ImageCLEF, demonstrate the proposed approaches outperform the existing methods by a large margin, especially on the large-scale challenging dataset. The code and models are available at https://github.com/BIT-DA/GDCAN .
Molecules with di‐ or trichloromethyl group are widely existed in natural products and man‐made structures that represent special activities in pharmaceuticals and agrochemicals. Besides, ...polychloromethyl‐containing compounds are also high valuable intermediates in organic synthesis and also can be used as functional materials. Therefore, significant efforts have been made toward the incorporation of polychloromethyl groups into certain structures. In this field, a variety of polychloromethyl reagents, such as CH2Cl2, HCCl3, CCl4, TMSCCl3, TMSCCl2H, etc. have been utilized as the di‐ or trichloromethyl sources toward polychloromethylated structures through radical, nucleophilic, and other reaction patterns. In this review, we summarized and discussed the recent achievements in the synthesis of polychloromethylated compounds, mainly focused on the employment of different polychloromethyl reagents. Most of the works we investigated were reported after 2010. Mechanism discussions and outlooks were also given.
Laser remelting (LR) is often used during selective laser melting (SLM) processes to improve the densification degree and top surface quality of the products. However, researches regarding its ...effects on side surface quality, residual stress, microstructure, and mechanical property are still quite lacking. The influence of the repeated usage of LR on a solidified layer is also not very clear. To address these issues, LR treatments with the cyclic numbers of 1–3 on every solidified layer have been employed during the SLM processes of a Ti-5Al-2.5Sn alloy in this study and their influences on deposition quality, residual stress, microstructure, and mechanical property were researched. The results first indicated that although the improvement in top surface quality and densification degree would be more and more notable with the increase of LR cycles, the side surface quality could not be improved through all the LR treatments. Then, it was shown by the hole-drilling tests on the surface centers of several 15 × 15 × 3 mm3 thin-plate specimens that when 1 cycle of LR was conducted on every solidified layer, the principal residual stress at the hole bottom of the SLM sample could be far beyond that of the non-treated one. In contrast, when 2 or 3 cycles of LR were conducted, the measured principal residual stresses could be reduced to lower than that of the non-treated ones. It was also proved that unlike the non-treated sample which presented a martensite-dominated microstructure with a very weak texture, all the LR-treated samples exhibited two kinds of preferential orientations. Finally, tensile properties of the stress-relieved samples were tested and it was seen that under the combined influence of the densification and the microstructure factors, the elongations of the LR-treated samples became slightly higher than that of the non-treated one whereas the tensile strengths were nearly identical. In summary, this paper demonstrates that there is still limitation in using LR as an auxiliary process of the SLM production of titanium products. The possible increase in residual stress and the formation of strong textures should be particularly considered.
•Laser remelting (LR) led to texture formation within the SLM Ti-5Al-2.5Sn.•Improvement in density and top surface quality is notable at higher LR cycles.•Imposing 1 cycle of LR on each solidified layer led to increase in residual stress.•Elongation increased after LR treatment whereas tensile strength was not affected.•Side surface quality could not be effectively improved through LR treatments.