APMG: 3D Molecule Generation Driven by Atomic Chemical Properties.

Recently, mask-fill-based 3D Molecular Generation (MG) methods have become very popular in virtual drug design. However, the existing MG methods ignore the chemical properties of atoms and contain inappropriate atomic position training data, which limits their generation capability. To mitigate the above issues, this paper presents a novel mask-fill-based 3D molecule generation model driven by atomic chemical properties (APMG). Specifically, we construct a new attention-MPNN-based encoder and introduce the electronic information into atom representations to enrich chemical properties. Also, a multi-functional classifier is designed to predict the electronic information of each generated atom, guiding the type prediction of elements and bonds. By design, the proposed method uses the chemical properties of atoms and their correlations for high-quality molecule generation. Second, to optimize the atomic position training data, we propose a novel atomic training position generation approach using the Chi-Square distribution. We evaluate our APMG method on the CrossDocked dataset and visualize the docking states of the pockets and generated molecules. The obtained results demonstrate the superiority and merits of APMG over the state-of-the-art approaches. The dataset and codes will be available on the project homepage: https://github.com/JU-HuaY/APMG.

Adaptive Log-Euclidean Metrics for SPD Matrix Learning.

Symmetric Positive Definite (SPD) matrices have received wide attention in machine learning due to their intrinsic capacity to encode underlying structural correlation in data. Many successful Riemannian metrics have been proposed to reflect the non-Euclidean geometry of SPD manifolds. However, most existing metric tensors are fixed, which might lead to sub-optimal performance for SPD matrix learning, especially for deep SPD neural networks. To remedy this limitation, we leverage the commonly encountered pullback techniques and propose Adaptive Log-Euclidean Metrics (ALEMs), which extend the widely used Log-Euclidean Metric (LEM). Compared with the previous Riemannian metrics, our metrics contain learnable parameters, which can better adapt to the complex dynamics of Riemannian neural networks with minor extra computations. We also present a complete theoretical analysis to support our ALEMs, including algebraic and Riemannian properties. The experimental and theoretical results demonstrate the merit of the proposed metrics in improving the performance of SPD neural networks. The efficacy of our metrics is further showcased on a set of recently developed Riemannian building blocks, including Riemannian batch normalization, Riemannian Residual blocks, and Riemannian classifiers.

Photothermal Catalytic Reduction and Bone Tissue Engineering Towards a Three-in-One Therapy Strategy for Osteosarcoma.

Osteosarcoma is one of the most dreadful bone neoplasms in young people, necessitating the development of innovative therapies that can effectively eliminate tumors while minimizing damage to limb function. An ideal therapeutic strategy should possess three essential capabilities: antitumor effects, tissue-protective properties, and the ability to enhance osteogenesis. In this study, self-assembled Ce-substituted molybdenum blue (CMB) nanowheel crystals are synthesized and loaded onto 3D-printed bioactive glass (CMB@BG) scaffolds to develop a unique three-in-one treatment approach for osteosarcoma. The CMB@BG scaffolds exhibit outstanding photothermally derived tumor ablation within the near-infrared-II window due to the surface plasmon resonance properties of the CMB nanowheel crystals. Furthermore, the photothermally synergistic catalytic effect of CMB promotes the rapid scavenging of reactive oxygen species caused by excessive heat, thereby suppressing inflammation and protecting surrounding tissues. The CMB@BG scaffolds possess pro-proliferation and pro-differentiation capabilities that efficiently accelerate bone regeneration within bone defects. Altogether, the CMB@BG scaffolds that combine highly efficient tumor ablation, tissue protection based on anti-inflammatory mechanisms, and enhanced osteogenic ability are likely to be a point-to-point solution for the comprehensive therapeutic needs of osteosarcoma.

Microneedle-Delivered PDA@Exo for Multifaceted Osteoarthritis Treatment via PI3K-Akt-mTOR Pathway.

Osteoarthritis (OA) is marked by cartilage deterioration, subchondral bone changes, and an inflammatory microenvironment. The study introduces the Microneedle-Delivered Polydopamine-Exosome (PDA@Exo MN), a therapeutic that not only preserves cartilage and promotes bone regeneration but also improves localized drug delivery through enhanced penetration capabilities. PDA@Exo MN shows strong reactive oxygen species (ROS) scavenging abilities and high biocompatibility, fostering osteogenesis and balancing anabolic and catabolic processes in cartilage. It directs macrophage polarization from M0 to the anti-inflammatory M2 phenotype. RNA sequencing of treated chondrocytes demonstrates restored cellular function and activated antioxidant responses, with modulated inflammatory pathways. The PI3K-AKT-mTOR pathway's activation, essential for PDA@Exo's effects, is confirmed via bioinformatics and Western blot. In vivo assessments robustly validate that PDA@Exo MN prevents cartilage degradation and OA progression, supported by histological assessments and micro-CT analysis, highlighting its disease-modifying impact. The excellent biocompatibility of PDA@Exo MN, verified through histological (H&E) and blood tests showing no organ damage, underscores its safety and efficacy for OA therapy, making it a novel and multifunctional nanomedical approach in orthopedics, characterized by organ-friendliness and biosecurity.

View-shuffled clustering via the modified Hungarian algorithm.

In the majority of existing multi-view clustering methods, the prerequisite is that the data have the correct cross-view correspondence. However, this strong assumption may not always hold in real-world applications, giving rise to the so-called View-shuffled Problem (VsP). To address this challenge, we propose a novel multi-view clustering method, namely View-shuffled Clustering via the Modified Hungarian Algorithm (VsC-mH). Specifically, we first establish the cross-view correspondence of the shuffled data utilizing strategies of the global alignment and modified Hungarian algorithm (mH) based intra-category alignment. Subsequently, we generate the partition of the aligned data employing matrix factorization. The fusion of these two processes facilitates the interaction of information, resulting in improved quality of both data alignment and partition. VsC-mH is capable of handling the data with alignment ratios ranging from 0 to 100%. Both experimental and theoretical evidence guarantees the convergence of the proposed optimization algorithm. Extensive experimental results obtained on six practical datasets demonstrate the effectiveness and merits of the proposed method.

Blocking CXCR4-CARM1-YAP axis overcomes osteosarcoma doxorubicin resistance by suppressing aerobic glycolysis.

Osteosarcoma, recognized for its aggressiveness and resistance to chemotherapy, notably doxorubicin, poses significant treatment challenges. This comprehensive study investigated the CXCR4-CARM1-YAP signaling axis and its pivotal function in controlling aerobic glycolysis, which plays a crucial role in doxorubicin resistance. Detailed analysis of Dox-resistant 143b/MG63-DoxR cells has uncovered the overexpression of CXCR4. Utilizing a combination of molecular biology techniques including gene silencing, aerobic glycolysis assays such as Seahorse experiments, RNA sequencing, and immunofluorescence staining. The study provides insight into the mechanistic pathways involved. Results demonstrated that disrupting CXCR4 expression sensitizes cells to doxorubicin-induced apoptosis and alters glycolytic activity. Further RNA sequencing revealed that CARM1 modulated this effect through its influence on glycolysis, with immunofluorescence of clinical samples confirming the overexpression of CXCR4 and CARM1 in drug-resistant tumors. Chromatin immunoprecipitation studies further highlighted the role of CARM1, showing it to be regulated by methylation at the H3R17 site, which in turn affected YAP expression. Crucially, in vivo experiments illustrated that CARM1 overexpression could counteract the tumor growth suppression that resulted from CXCR4 inhibition. These insights revealed the intricate mechanisms at play in osteosarcoma resistance to doxorubicin and pointed toward potential new therapeutic strategies that could target this metabolic and signaling network to overcome drug resistance and improve patient outcomes.

Modeling blood-brain barrier formation and cerebral cavernous malformations in human PSC-derived organoids.

The human blood-brain barrier (hBBB) is a highly specialized structure that regulates passage across blood and central nervous system (CNS) compartments. Despite its critical physiological role, there are no reliable in vitro models that can mimic hBBB development and function. Here, we constructed hBBB assembloids from brain and blood vessel organoids derived from human pluripotent stem cells. We validated the acquisition of blood-brain barrier (BBB)-specific molecular, cellular, transcriptomic, and functional characteristics and uncovered an extensive neuro-vascular crosstalk with a spatial pattern within hBBB assembloids. When we used patient-derived hBBB assembloids to model cerebral cavernous malformations (CCMs), we found that these assembloids recapitulated the cavernoma anatomy and BBB breakdown observed in patients. Upon comparison of phenotypes and transcriptome between patient-derived hBBB assembloids and primary human cavernoma tissues, we uncovered CCM-related molecular and cellular alterations. Taken together, we report hBBB assembloids that mimic the core properties of the hBBB and identify a potentially underlying cause of CCMs.

Temperature-Controlled Selective Formation of Silver Nanoclusters and Their Transformation to the Same Product.

Herein, two atomically precise silver nanoclusters, Ag54 and Ag33, directed by inner anion templates (CrO4 2- and/or Cl-), are initially isolated as a mixed phase from identical reactants across a wide temperature range (20-80 °C). Interestingly, fine-tuning the reaction temperature can realize pure phase synthesis of the two nanoclusters; that is, a metastable Ag54 is kinetically formed at a low temperature (20 °C), whereas such a system is steered towards a thermodynamically stable Ag33 at a relatively high temperature (80 °C). Electrospray ionization mass spectrometry illustrates that the stability of Ag33 is superior to that of Ag54, which is further supported by density functional theory calculations. Importantly, the difference in structural stability can influence the pathway of 1,4-bis(pyrid-4-yl)benzene induced transformation reaction starting from Ag54 and Ag33. The former undergoes a dramatic breakage-reorganization process to form an Ag31 dimer (Ag31), while the same product can be also achieved from the latter following a noninvasive ligand exchange process. Both the Ag54 and Ag33 have the potential for further remote laser ignition applications. This work not only demonstrates how temperature controls the isolation of a specific phase, but also sheds light on the structural transformation pathway of nanoclusters with different stability.

Depression heightened the association of the systemic immune-inflammation index with all-cause mortality among osteoarthritis patient.

BACKGROUND: Systemic immune-inflammatory index (SII) has been recognized as a novel inflammatory indicator in numerous diseases. It remains unknown how SII affects all-cause mortality among patients with osteoarthritis (OA). In this prospective cohort study, we intended to examine the relationship of SII with all-cause mortality among OA populations and assess the interaction between depression and SII. METHODS: Data was collected from National Health and Nutrition Examination Survey (NHANES) in 2005-2018. The National Death Index (NDI) provided vital status records. Multivariable Cox regression analyses with cubic spines were applied to estimate the association between SII and all-cause and CVD mortality. Stratified analysis and interaction tests assessed the interaction of SII and depression on all-cause mortality. RESULTS: In total 3174 OA adults were included. The lowest quartile Q1 (HR:1.44, 95%CI:1.02-2.04) and highest quartile Q4 (HR:1.44, 95%CI:1.02-2.04) of SII presented a higher risk of death compared with those in second quartile Q2 (Ref.) and third quartile Q3 (HR:1.23, 95%CI:0.89-1.68. Restricted cubic splines analysis revealed a U-shaped association of SII with all-cause mortality, the inflection points were 412.93 × 109/L. The interaction test observed a more significant relationship of SII with all-cause mortality in depression patients than in non-depression patients, indicating that depression can modify this association. LIMITATIONS: First, the observational study design failed to make causal inferences. Second, the baseline SII cannot reflect the long-term level of inflammation. Finally, there may be potential bias. CONCLUSION: SII was U-shaped associated with all-cause mortality in OA patients, and this association was significantly heightened by depression.

Oligosaccharides from black ginseng innovatively prepared by low-temperature steam-heating process ameliorate cognitive impairment in Alzheimer's disease mice via the Keap-1/Nrf2 pathway.

BACKGROUND: Our objective in this study was to evaluate the effectiveness of oligosaccharides extracted from black ginseng (OSBG), innovatively prepared by a low-temperature steam-heating process, in the improvement of learning and memory impairment in mice, as well as the mechanism(s). RESULTS: Eight carbohydrates involving isomaltose and maltotetraose were detected in black gensing; monosaccharide residues including mannose and rhamnose were also discovered. OSBG-treated mice showed significant amelioration in recognition and spatial memory deficits compared to the scopolamine group. OSBG could decrease acetylcholinesterase activity in a tissue-dependent fashion but not in a dose-dependent manner. Furthermore, in contrast, OSBG administration resulted in significant upregulation superoxide dismutase, glutathione, glutathione peroxidase (GPx), and Kelch-like ECH-associated protein 1, downregulation of malondialdehyde and nuclear factor erythroid 2-related factor 2 in the tissues. Finally, at the genus level, we observed that the OSBG interventions increased the relative abundance of probiotics (e.g., Barnesiella, Staphylococcus, Clostridium_XlVb) and decreased pernicious bacteria such as Eisenbergiella and Intestinimonas, compared to the Alzheimer's disease mouse model group. Herein, our results demonstrate that OSBG restores the composition of the scopolamine-induced intestinal microbiota in mice, providing homeostasis of gut microbiota and providing evidence for microbiota-regulated therapeutic potential. CONCLUSION: Our results showed for the first time a clear role for OSBG in improving scopolamine-induced memory impairment by inhibiting cholinergic dysfunction in a tissue-dependent manner. Additionally, OSBG administration relieved oxidative stress by activating the Keap-1/Nrf2 pathway and modulating the gut microbiota. Collectively, OSBG may be a promising target for neuroprotective antioxidants for improving memory and cognition in Alzheimer's disease patients. © 2024 Society of Chemical Industry.

LINC00960 affects osteosarcoma treatment and prognosis by regulating the tumor immune microenvironment.

Background: Osteosarcoma (OS), the commonest primary malignant bone tumor, is mainly seen in children and teenagers. LINC00960, a newly discovered long intergenic non-protein coding RNA, has been shown to be important in certain cancers. The objective of this study was to assess LINC00960's prognostic and therapeutic value and analyze its mechanism of action in osteosarcoma. Methods: With the transcriptome information of 85 osteosarcomas from the TARGET database, the Cox regression analyses, K-M curve, and ROC curve, were conducted for survival and prognostic analysis. The functional analysis was conducted using GO, KEGG, GSEA, and GSVA. The ESTIMATE, ssGSEA, MCP-counter, ImmuCellAI algorithms, and immune checkpoint correlation analysis were performed for immune-related analysis. The single-cell RNA sequencing data of 6 osteosarcoma patients was obtained from the Gene Expression Omnibus database. The Tumor Immune Dysfunction and Exclusion algorithm and the "pRRophetic" R package were performed to predict the response to immunotherapy and chemotherapy. Results: LINC00960 overexpression is associated with osteosarcoma metastasis and poor prognosis. Based on the LINC00960 expression, the nomogram prediction model was created, which showed good accuracy and precision to predict the overall survival of osteosarcoma. Single-cell and immune-related analysis showed that LINC00960 is mainly highly expressed in the tumor-exhausted CD8 T cells in osteosarcoma. In osteosarcoma, the expression of LIC00960 was favorably connected with immune checkpoint-related genes and negatively correlated with immune infiltration. TIDE analysis indicated that low LINC00960 expression patients might have a better response to immunotherapy. Drug sensitivity analysis showed that high LINC00960 expression patients might have better responses to Bleomycin and Doxorubicin. Conclusion: LINC00960 has the potential to be a novel biomarker for predicting overall survival in osteosarcoma patients and to guide more individualized treatment and clinical decision-making.

Soil microbiota plays a key regulatory role in the outbreak of tobacco root rot.

Introduction: Root rot caused by the fungal pathogen Fusarium sp. poses significant challenges to tobacco cultivation in China, leading to major economic setbacks. The interplay between this pathogen and the wider soil microbial community remains poorly understood. Methods: High-throughput sequencing technology was utilized to evaluate soil prokaryotic, fungal, and protistan communities. We compared microbial communities in infected soils to those in healthy soils from the same field. Additionally, the influence of pH on the microbial communities was assessed. Results: Infected soils displayed elevated levels of soil nutrients but diminished observed richness across prokaryotic, fungal, and protistan groups. The pathogenic fungi Fusarium solani f sp. eumartii's abundance was notably increased in infected soils. Infection with F. solani significantly altered the soil's microbial community structure and interactions, manifested as a decrease in network scale and the number of keystone species. An evaluation of prokaryotes' role in F. solani's invasion revealed an increased number of connecting nodes in infected soils. Additionally, relationships between predatory protists and fungi were augmented, whereas predation on F. solani declined. Discussion: The study underscores the significance of comprehending the interactions among soil microorganisms and brings to light the susceptibility of soil microbial communities to pathogen invasion. It offers insights into the multifaceted relationships and potential vulnerabilities within the soil ecosystem in the context of Fusarium sp. invasion.

High-Resolution Characterization of Human Brain Cortex with High-Fidelity Spatial Transcriptomic Slides (HiFi-Slides).

Spatial transcriptomic tools and platforms help researchers to inspect tissues and cells with fine details of how they differentiate in expressions and how they orient themselves. With the higher resolution we get and higher throughput of expression targets, spatial analysis can truly become the core player for cell clustering, migration study, and, eventually, the novel model for pathological study. We present the demonstration of HiFi-slide, a whole transcriptomic sequencing technique that recycles used sequenced-by-synthesis flow cell surfaces to a high-resolution spatial mapping tool that can be directly applied to tissue cell gradient analysis, gene expression analysis, cell proximity analysis, and other cellular-level spatial studies.

Identification of 3 key genes as novel diagnostic and therapeutic targets for OA and COVID-19.

Background: Corona Virus Disease 2019 (COVID-19) and Osteoarthritis (OA) are diseases that seriously affect the physical and mental health and life quality of patients, particularly elderly patients. However, the association between COVID-19 and osteoarthritis at the genetic level has not been investigated. This study is intended to analyze the pathogenesis shared by OA and COVID-19 and to identify drugs that could be used to treat SARS-CoV-2-infected OA patients. Methods: The four datasets of OA and COVID-19 (GSE114007, GSE55235, GSE147507, and GSE17111) used for the analysis in this paper were obtained from the GEO database. Common genes of OA and COVID-19 were identified through Weighted Gene Co-Expression Network Analysis (WGCNA) and differential gene expression analysis. The least absolute shrinkage and selection operator (LASSO) algorithm was used to screen key genes, which were analyzed for expression patterns by single-cell analysis. Finally, drug prediction and molecular docking were carried out using the Drug Signatures Database (DSigDB) and AutoDockTools. Results: Firstly, WGCNA identified a total of 26 genes common between OA and COVID-19, and functional analysis of the common genes revealed the common pathological processes and molecular changes between OA and COVID-19 are mainly related to immune dysfunction. In addition, we screened 3 key genes, DDIT3, MAFF, and PNRC1, and uncovered that key genes are possibly involved in the pathogenesis of OA and COVID-19 through high expression in neutrophils. Finally, we established a regulatory network of common genes between OA and COVID-19, and the free energy of binding estimation was used to identify suitable medicines for the treatment of OA patients infected with SARS-CoV-2. Conclusion: In the present study, we succeeded in identifying 3 key genes, DDIT3, MAFF, and PNRC1, which are possibly involved in the development of both OA and COVID-19 and have high diagnostic value for OA and COVID-19. In addition, niclosamide, ciclopirox, and ticlopidine were found to be potentially useful for the treatment of OA patients infected with SARS-CoV-2.

Superior capsular reconstruction using the long head of the biceps to treat massive rotator cuff tears improves patients shoulder pain, mobility and function.

PURPOSE: Arthroscopic superior capsule reconstruction (SCR) with the long head of the biceps (LHBT) was performed to restore structural stability, force couple balance, and shoulder joint function. This study aimed to evaluate the functional outcomes of SCR using the LHBT over at least 24 months of follow-up. METHOD: This retrospective study included 89 patients with massive rotator cuff tears who underwent SCR using the LHBT, met the inclusion criteria and underwent follow up for at least 24 months. The preoperative and postoperative shoulder range of motion (forward flexion, external rotation, and abduction), acromiohumeral interval (AHI), visual analog scale (VAS) score, American Shoulder and Elbow Surgeons (ASES) score and Constant-Murley score were obtained, and the tear size, and Goutallier and Hamada grades were also investigated. RESULTS: Compared with those measured preoperatively, the range of motion, AHI, and VAS, Constant-Murley, and ASES scores were significantly improved immediately postoperatively (P < 0.001) and at the 6-month, 12-month, and final follow-ups (P < 0.001). At the last follow-up, the postoperative ASES score and Constant-Murley score increased from 42.8 ± 7.6 to 87.4 ± 6.1, and 42.3 ± 8.9 to 84.9 ± 10.7, respectively; with improvements of 51 ± 21.7 in forward flexion, 21.0 ± 8.1 in external rotation, and 58.5 ± 22.5 in abduction. The AHI increased 2.1 ± 0.8 mm and the VAS score significantly changed from 6.0 (5.0, 7.0) to 1.0 (0.0, 1.0), at the final follow-up. Eleven of the 89 patients experienced retears, and one patient needed reoperation. CONCLUSION: In this study with at least 24-months of follow-up, SCR using the LHBT for massive rotator cuff tears could effectively relieve shoulder pain, restore shoulder function and increase shoulder mobility to some extent. LEVEL OF EVIDENCE: IV.

TGFuse: An Infrared and Visible Image Fusion Approach Based on Transformer and Generative Adversarial Network.

The end-to-end image fusion framework has achieved promising performance, with dedicated convolutional networks aggregating the multi-modal local appearance. However, long-range dependencies are directly neglected in existing CNN fusion approaches, impeding balancing the entire image-level perception for complex scenario fusion. In this paper, therefore, we propose an infrared and visible image fusion algorithm based on the transformer module and adversarial learning. Inspired by the global interaction power, we use the transformer technique to learn the effective global fusion relations. In particular, shallow features extracted by CNN are interacted in the proposed transformer fusion module to refine the fusion relationship within the spatial scope and across channels simultaneously. Besides, adversarial learning is designed in the training process to improve the output discrimination via imposing competitive consistency from the inputs, reflecting the specific characteristics in infrared and visible images. The experimental performance demonstrates the effectiveness of the proposed modules, with superior improvement against the state-of-the-art, generalising a novel paradigm via transformer and adversarial learning in the fusion task.

LRRNet: A Novel Representation Learning Guided Fusion Network for Infrared and Visible Images.

Deep learning based fusion methods have been achieving promising performance in image fusion tasks. This is attributed to the network architecture that plays a very important role in the fusion process. However, in general, it is hard to specify a good fusion architecture, and consequently, the design of fusion networks is still a black art, rather than science. To address this problem, we formulate the fusion task mathematically, and establish a connection between its optimal solution and the network architecture that can implement it. This approach leads to a novel method proposed in the paper of constructing a lightweight fusion network. It avoids the time-consuming empirical network design by a trial-and-test strategy. In particular we adopt a learnable representation approach to the fusion task, in which the construction of the fusion network architecture is guided by the optimisation algorithm producing the learnable model. The low-rank representation (LRR) objective is the foundation of our learnable model. The matrix multiplications, which are at the heart of the solution are transformed into convolutional operations, and the iterative process of optimisation is replaced by a special feed-forward network. Based on this novel network architecture, an end-to-end lightweight fusion network is constructed to fuse infrared and visible light images. Its successful training is facilitated by a detail-to-semantic information loss function proposed to preserve the image details and to enhance the salient features of the source images. Our experiments show that the proposed fusion network exhibits better fusion performance than the state-of-the-art fusion methods on public datasets. Interestingly, our network requires a fewer training parameters than other existing methods.

Discriminative Dictionary Pair Learning With Scale-Constrained Structured Representation for Image Classification.

The dictionary pair learning (DPL) model aims to design a synthesis dictionary and an analysis dictionary to accomplish the goal of rapid sample encoding. In this article, we propose a novel structured representation learning algorithm based on the DPL for image classification. It is referred to as discriminative DPL with scale-constrained structured representation (DPL-SCSR). The proposed DPL-SCSR utilizes the binary label matrix of dictionary atoms to project the representation into the corresponding label space of the training samples. By imposing a non-negative constraint, the learned representation adaptively approximates a block-diagonal structure. This innovative transformation is also capable of controlling the scale of the block-diagonal representation by enforcing the sum of within-class coefficients of each sample to 1, which means that the dictionary atoms of each class compete to represent the samples from the same class. This implies that the requirement of similarity preservation is considered from the perspective of the constraint on the sum of coefficients. More importantly, the DPL-SCSR does not need to design a classifier in the representation space as the label matrix of the dictionary can also be used as an efficient linear classifier. Finally, the DPL-SCSR imposes the l2,p -norm on the analysis dictionary to make the process of feature extraction more interpretable. The DPL-SCSR seamlessly incorporates the scale-constrained structured representation learning, within-class similarity preservation of representation, and the linear classifier into one regularization term, which dramatically reduces the complexity of training and parameter tuning. The experimental results on several popular image classification datasets show that our DPL-SCSR can deliver superior performance compared with the state-of-the-art (SOTA) dictionary learning methods. The MATLAB code of this article is available at https://github.com/chenzhe207/DPL-SCSR.

Towards Robust Visual Object Tracking with Independent Target-Agnostic Detection and Effective Siamese Cross-Task Interaction.

Advanced Siamese visual object tracking architectures are jointly trained using pair-wise input images to perform target classification and bounding box regression. They have achieved promising results in recent benchmarks and competitions. However, the existing methods suffer from two limitations: First, though the Siamese structure can estimate the target state in an instance frame, provided the target appearance does not deviate too much from the template, the detection of the target in an image cannot be guaranteed in the presence of severe appearance variations. Second, despite the classification and regression tasks sharing the same output from the backbone network, their specific modules and loss functions are invariably designed independently, without promoting any interaction. Yet, in a general tracking task, the centre classification and bounding box regression tasks are collaboratively working to estimate the final target location. To address the above issues, it is essential to perform target-agnostic detection so as to promote cross-task interactions in a Siamese-based tracking framework. In this work, we endow a novel network with a target-agnostic object detection module to complement the direct target inference, and to avoid or minimise the misalignment of the key cues of potential template-instance matches. To unify the multi-task learning formulation, we develop a cross-task interaction module to ensure consistent supervision of the classification and regression branches, improving the synergy of different branches. To eliminate potential inconsistencies that may arise within a multi-task architecture, we assign adaptive labels, rather than fixed hard labels, to supervise the network training more effectively. The experimental results obtained on several benchmarks, i.e., OTB100, UAV123, VOT2018, VOT2019, and LaSOT, demonstrate the effectiveness of the advanced target detection module, as well as the cross-task interaction, exhibiting superior tracking performance as compared with the state-of-the-art tracking methods.

Fast Self-Guided Multi-View Subspace Clustering.

Multi-view subspace clustering is an important topic in cluster analysis. Its aim is to utilize the complementary information conveyed by multiple views of objects to be clustered. Recently, view-shared anchor learning based multi-view clustering methods have been developed to speed up the learning of common data representation. Although widely applied to large-scale scenarios, most of the existing approaches are still faced with two limitations. First, they do not pay sufficient consideration on the negative impact caused by certain noisy views with unclear clustering structures. Second, many of them only focus on the multi-view consistency, yet are incapable of capturing the cross-view diversity. As a result, the learned complementary features may be inaccurate and adversely affect clustering performance. To solve these two challenging issues, we propose a Fast Self-guided Multi-view Subspace Clustering (FSMSC) algorithm which skillfully integrates the view-shared anchor learning and global-guided-local self-guidance learning into a unified model. Such an integration is inspired by the observation that the view with clean clustering structures will play a more crucial role in grouping the clusters when the features of all views are concatenated. Specifically, we first learn a locally-consistent data representation shared by all views in the local learning module, then we learn a globally-discriminative data representation from multi-view concatenated features in the global learning module. Afterwards, a feature selection matrix constrained by the l2,1 -norm is designed to construct a guidance from global learning to local learning. In this way, the multi-view consistent and diverse information can be simultaneously utilized and the negative impact caused by noisy views can be overcame to some extent. Extensive experiments on different datasets demonstrate the effectiveness of our proposed fast self-guided learning model, and its promising performance compared to both, the state-of-the-art non-deep and deep multi-view clustering algorithms. The code of this paper is available at https://github.com/chenzhe207/FSMSC.