Suboptimal Results Research Articles

OGNNMDA: a computational model for microbe-drug association prediction based on ordered message-passing graph neural networks.

In recent years, many excellent computational models have emerged in microbe-drug association prediction, but their performance still has room for improvement. This paper proposed the OGNNMDA framework, which applied an ordered message-passing mechanism to distinguish the different neighbor information in each message propagation layer, and it achieved a better embedding ability through deeper network layers. Firstly, the method calculates four similarity matrices based on microbe functional similarity, drug chemical structure similarity, and their respective Gaussian interaction profile kernel similarity. After integrating these similarity matrices, it concatenates the integrated similarity matrix with the known association matrix to obtain the microbe-drug heterogeneous matrix. Secondly, it uses a multi-layer ordered message-passing graph neural network encoder to encode the heterogeneous network and the known association information adjacency matrix, thereby obtaining the final embedding features of the microbe-drugs. Finally, it inputs the embedding features into the bilinear decoder to get the final prediction results. The OGNNMDA method performed comparative experiments, ablation experiments, and case studies on the aBiofilm, MDAD and DrugVirus datasets using 5-fold cross-validation. The experimental results showed that OGNNMDA showed the strongest prediction performance on aBiofilm and MDAD and obtained sub-optimal results on DrugVirus. In addition, the case studies on well-known drugs and microbes also support the effectiveness of the OGNNMDA method. Source codes and data are available at: https://github.com/yyzg/OGNNMDA.

The Multiobjective Control Based on Tolerance Optimization in a Multienergy System

To address the issue of multiobjective control in multienergy systems with diverse operational objectives, a two-stage optimization framework based on expected point tolerance has been proposed in this paper. In the first stage, a single objective function is used for optimization control to obtain the expected point of the multiobjective optimization problem. Then, in the second stage, by defining the allowable deviation between each optimization objective and the expected point, the original multiobjective optimization problem is transformed into a single objective optimization problem solution with tolerance measurement. Finally, in the simulation scene of a multienergy system, it is demonstrated that compared with the optimal results under each single objective method, the proposed method increases power line loss, maximum voltage deviation, new energy consumption, and economy by 2.22, 2.30, 1.02, and 2.45 times, respectively. Compared with the suboptimal results, the proposed method reduces power line loss by 22.26, 1.74, 1.09, and 0.97 times, respectively. Combining the shape of the Pareto frontier, it is demonstrated that the proposed method can comprehensively consider the needs of multiple power optimization objectives for forming a more reasonable and effective system optimization scheduling and also provide a new approach for solving multiobjective optimization problems.

From omnivores to vegans: Differences and similitudes among different dietary patterns in Spain. Nutrition knowledge, dietary recommendations compliance and level of food supplementation

Western diets are associated with the rise in prevalent diseases of our time and have a significant environmental impact. The growing concern over these issues has led to a trend of shifting away from traditional omnivorous diets towards various dietary patterns with differing levels of animal-origin food restriction, such as flexitarian, pescatarian, vegetarian, or vegan diets. While much of the current literature has focused on analysing the motivations and barriers of these dietary groups, little is known about other aspects that may unite or differentiate them beyond their food intake preferences. In 2022, a survey was conducted in Spain, where 38% of the population follows a non-omnivorous diet. A total of 792 participants, with equal representation across omnivores and vegans, took part in the survey. The objective was to identify various aspects related to different dietary patterns, including knowledge, attitudes, and perceptions towards food. This study presents novel findings regarding knowledge of healthy eating, adherence to national dietary recommendations, and levels of food supplementation. Approximately half of the participants, regardless of dietary pattern, were found to be unaware of the composition of a healthy and balanced meal. Omnivores demonstrated the highest compliance with recommendations for dairy and protein intake but showed poorer adherence to fruit and vegetable recommendations. Plant-based diet groups also displayed suboptimal results. Both vegetarians and vegans exhibited higher rates of supplementation, though not as high as expected. Overall, there appears to be a widespread lack of knowledge among the Spanish population regarding healthy and balanced diets. The prevalent non-compliance with dietary recommendations and the level of dietary supplementation among the more restrictive dietary groups, suggest potential inadequacies in diet quality within these populations. These findings can inform future interventions aimed at improving dietary patterns among these population groups in terms of health, guidance, or food development.

CCSS: Towards conductance-based community search with size constraints

Size-constrained community search, retrieving a size-bounded high-quality subgraph containing user-specified query vertices, has been extensively studied in graph analysis. However, existing methods mainly focus on the cohesiveness within the community while ignoring its separation from the outside, leading to sub-optimal results. Inspired by this, we adopt the well-known conductance to measure the quality of a community and introduce a novel problem of Conductance-based Community Search with Size Constraints (CCSS). CCSS aims to identify a subgraph with the smallest conductance among all connected subgraphs that contain the query vertex q and have at least l and at most h vertices. We prove that CCSS is NP-hard. To efficiently address CCSS, we first propose a computational framework ISC with two-level loops. In the inner loop, ISC starts from q and iteratively adds vertices based on the predefined score function, obtaining a coarse-grained candidate community. In the outer loop, we design the perturbation strategy to repeatedly execute the inner loop until the quality of the candidate community converges. Then, we propose two heuristic algorithms, ISCCI and ISCCP, to concretely implement the score function and perturbation strategy of ISC, which significantly reduce the computational costs. Finally, empirical results demonstrate the superiorities of our solutions. Our source codes and datasets are available at https://github.com/longlonglin/CCSS.

SSGCL: Simple Social Recommendation with Graph Contrastive Learning

As user–item interaction information is typically limited, collaborative filtering (CF)-based recommender systems often suffer from the data sparsity issue. To address this issue, recent recommender systems have turned to graph neural networks (GNNs) due to their superior performance in capturing high-order relationships. Furthermore, some of these GNN-based recommendation models also attempt to incorporate other information. They either extract self-supervised signals to mitigate the data sparsity problem or employ social information to assist with learning better representations under a social recommendation setting. However, only a few methods can take full advantage of these different aspects of information. Based on some testing, we believe most of these methods are complex and redundantly designed, which may lead to sub-optimal results. In this paper, we propose SSGCL, which is a recommendation system model that utilizes both social information and self-supervised information. We design a GNN-based propagation strategy that integrates social information with interest information in a simple yet effective way to learn user–item representations for recommendations. In addition, a specially designed contrastive learning module is employed to take advantage of the self-supervised signals for a better user–item representation distribution. The contrastive learning module is jointly optimized with the recommendation module to benefit the final recommendation result. Experiments on several benchmark data sets demonstrate the significant improvement in performance achieved by our model when compared with baseline models.

Token-disentangling Mutual Transformer for multimodal emotion recognition

Multimodal emotion recognition presents a complex challenge, as it involves the identification of human emotions using various modalities such as video, text, and audio. Existing methods focus mainly on the fusion information from multimodal data, but ignore the interaction of the modality-specific heterogeneity features that contribute differently to emotions, leading to sub-optimal results. To tackle this challenge, we propose a novel Token-disentangling Mutual Transformer (TMT) for robust multimodal emotion recognition, by effectively disentangling and interacting inter-modality emotion consistency features and intra-modality emotion heterogeneity features. Specifically, the TMT consists of two main modules: multimodal emotion Token disentanglement and Token mutual Transformer. In the multimodal emotion Token disentanglement, we introduce a Token separation encoder with an elaborated Token disentanglement regularization, which effectively disentangle the inter-modality emotion consistency feature Token from each intra-modality emotion heterogeneity feature Token; consequently, the emotion-related consistency and heterogeneity information can be performed independently and comprehensively. Furthermore, we devise the Token mutual Transformer with two cross-modal encoders to interact and fuse the disentangled feature Tokens by using bi-directional query learning, which delivers more comprehensive and complementary multimodal emotion representations for multimodal emotion recognition. We evaluate our model on three popular three-modality emotion datasets, namely CMU-MOSI, CMU-MOSEI, and CH-SIMS, and the experimental results affirm the superior performance of our model compared to state-of-the-art methods, achieving state-of-the-art recognition performance. Evaluation Codes and models are released at https://github.com/cug-ygh/TMT.

Effect of Focal Zone Size on Treatment Outcomes and Renal Injury Following Extracorporeal Shockwave Lithotripsy of Renal Calculi: A Prospective Randomized Study.

Background: The narrower focal zone (FZ) size of modern lithotripter was considered as one of the factors that resulted in suboptimal treatment result of extracorporeal shockwave lithotripsy (SWL). Therefore, we investigate the efficacy and safety of standard narrow or extended (FZ) sizes in SWL for patients with renal stones. Materials and Methods: In this prospective study conducted between April 2018 and October 2022, patients with renal stones were randomized to receive SWL with either standard or extended FZ. Treatment was delivered using a Modulith SLX-F2 lithotripter with a maximum of 3000 shocks at 1.5 Hz. The primary outcome was treatment success 12 weeks after a single SWL session, defined as the absence of a stone or stone fragment <4 mm on computed tomography. Secondary outcomes included the incidence of perinephric hematoma, stone-free rate (SFR), and changes in the urinary levels of acute renal injury markers. Results: A total of 320 patients were recruited, and 276 patients were randomized into the two groups. The two groups had similar baseline parameters. The treatment success rate was significantly better for standard FZ (74.3%) than the extended FZ group (59.3%) (p = 0.009). Standard FZ also had a significantly better SFR (Grade-A, 36.8% vs 23.0%, p = 0.013) and less pain after treatment. Both groups had similar perinephric hematoma formation rates, unplanned hospital admission rates, and changes in urinary acute renal injury markers. Conclusions: The standard narrow FZ has better treatment efficacy and similar safety compared with the extended FZ during SWL for renal stones. This clinical trial has been registered in the public domain (CCRBCTR) under trial number CUHK_CCRB00510.

Optimisation of hybrid composite shields for hypervelocity Micro-Meteoroid and Orbital Debris (MMOD) impact

Advancements in space exploration over the last few decades have led to a sharp increase in Micro-Meteoroid and Orbital Debris (MMOD), with the associated increased risk of hypervelocity impact on satellites and space structures. The Whipple shield can mitigate the effects of such impacts and current research is exploring further developments towards effective lightweight passive shielding technology to counter the damaging effects of MMOD. With the increase in MMOD and the prospect of increased space travel in the coming years and decades, it is vital that more research is conducted and improvements are made in the development and design of efficient, lightweight shielding technology to protect both unmanned and manned spacecraft against hypervelocity impacts (HVI). The optimisation of shield design for HVI is a high dimensionality problem, suited to advanced computational approaches. Key variables in HVI shield design that should inform the focus on optimisation include: impact velocity, rear wall thickness, projectile diameter and bumper thickness. A hybrid shield configuration and numerical model are proposed and validated, with alternating layers of aluminium (AL2024) and carbon fibre composites (CFRP, T300 woven-fabric) to form a 5 mm thick target plate consisting of 5 plies. The adaptive coupled FEM-SPH method is used to model the target plates. The proposed hypervelocity shield design optimisation methodology is based on Direct Simulation-based Genetic Algorithm (DSGA) optimisation and is implemented to optimise a multi-variable shield design space. Objective weightings are used to analyse and discuss results, referring to the ratio of the kinetic energy to the shield areal density objectives. A clear transition in the impact behaviour of the optimised MMOD shields is observed in the transition region from high-velocity to hypervelocity impact, where significant levels of kinetic energy dissipation are observed below the transition region, and lower energy dissipation at hypervelocity. The shield design optimisation results show that with a weighted kinetic energy to mass objective of 90:10, the kinetic energy of the back shield plate decreases significantly (62.7%) and the areal density can be reduced by more than 18%. Alternative configurations displayed sub-optimal results based on a trade-off between objective functions.

Deep Self-Supervised Attributed Graph Clustering for Social Network Analysis

Deep graph clustering is an unsupervised learning task that divides nodes in a graph into disjoint regions with the help of graph auto-encoders. Currently, such methods have several problems, as follows. (1) The deep graph clustering method does not effectively utilize the generated pseudo-labels, resulting in sub-optimal model training results. (2) Each cluster has a different confidence level, which affects the reliability of the pseudo-label. To address these problems, we propose a Deep Self-supervised Attribute Graph Clustering model (DSAGC) to fully leverage the information of the data itself. We divide the proposed model into two parts: an upstream model and a downstream model. In the upstream model, we use the pseudo-label information generated by spectral clustering to form a new high-confidence distribution with which to optimize the model for a higher performance. We also propose a new reliable sample selection mechanism to obtain more reliable samples for downstream tasks. In the downstream model, we only use the reliable samples and the pseudo-label for the semi-supervised classification task without the true label. We compare the proposed method with 17 related methods on four publicly available citation network datasets, and the proposed method generally outperforms most existing methods in three performance metrics. By conducting a large number of ablative experiments, we validate the effectiveness of the proposed method.

Online Bayesian optimization of vagus nerve stimulation.

Objective.In bioelectronic medicine, neuromodulation therapies induce neural signals to the brain or organs, modifying their function. Stimulation devices capable of triggering exogenous neural signals using electrical waveforms require a complex and multi-dimensional parameter space to control such waveforms. Determining the best combination of parameters (waveform optimization or dosing) for treating a particular patient's illness is therefore challenging. Comprehensive parameter searching for an optimal stimulation effect is often infeasible in a clinical setting due to the size of the parameter space. Restricting this space, however, may lead to suboptimal therapeutic results, reduced responder rates, and adverse effects.Approach. As an alternative to a full parameter search, we present a flexible machine learning, data acquisition, and processing framework for optimizing neural stimulation parameters, requiring as few steps as possible using Bayesian optimization. This optimization builds a model of the neural and physiological responses to stimulations, enabling it to optimize stimulation parameters and provide estimates of the accuracy of the response model. The vagus nerve (VN) innervates, among other thoracic and visceral organs, the heart, thus controlling heart rate (HR), making it an ideal candidate for demonstrating the effectiveness of our approach.Main results.The efficacy of our optimization approach was first evaluated on simulated neural responses, then applied to VN stimulation intraoperatively in porcine subjects. Optimization converged quickly on parameters achieving target HRs and optimizing neural B-fiber activations despite high intersubject variability.Significance.An optimized stimulation waveform was achieved in real time with far fewer stimulations than required by alternative optimization strategies, thus minimizing exposure to side effects. Uncertainty estimates helped avoiding stimulations outside a safe range. Our approach shows that a complex set of neural stimulation parameters can be optimized in real-time for a patient to achieve a personalized precision dosing.

Expression and functional characterization of chimeric recombinant bovine follicle-stimulating hormone produced in Leishmania tarentolae.

Assisted reproductive techniques are routinely used in livestock species to increase and enhance productivity. Ovarian hyperstimulation is a process that currently relies on administering pituitary-derived follicle-stimulating hormone (FSH) or equine chorionic gonadotropin in combination with other hormones to promote the maturation of multiple follicles and thereby achieve superovulation. The use of partially purified preparations of FSH extracted from natural sources is associated with suboptimal and variable results. Recombinant FSH (rFSH) has been produced in a variety of heterologous organisms. However, attaining a bioactive rFSH of high quality and at low cost for use in livestock remains challenging. Here we report the production and characterization of a single chain bovine rFSH consisting of the β- and α-subunit fused by a polypeptide linker (scbFSH) using Leishmania tarentolae as heterologous expression system. This unicellular eukaryote is non-pathogenic to mammals, can be grown in bioreactors using simple and inexpensive semisynthetic media at 26°C and does not require CO2 or bovine serum supplementation. Stable cell lines expressing scbFSH in an inducible fashion were generated and characterized for their productivity. Different culture conditions and purification procedures were evaluated, and the recombinant product was biochemically and biologically characterized, including bioassays in an animal model. The results demonstrate that L. tarentolae is a suitable host for producing a homogeneous, glycosylated and biologically active form of scbFSH with a reasonable yield.

Fast Multilabel Feature Selection via Global Relevance and Redundancy Optimization.

Information theoretical-based methods have attracted a great attention in recent years and gained promising results for multilabel feature selection (MLFS). Nevertheless, most of the existing methods consider a heuristic way to the grid search of important features, and they may also suffer from the issue of fully utilizing labeling information. Thus, they are probable to deliver a suboptimal result with heavy computational burden. In this article, we propose a general optimization framework global relevance and redundancy optimization (GRRO) to solve the learning problem. The main technical contribution in GRRO is a formulation for MLFS while feature relevance, label relevance (i.e., label correlation), and feature redundancy are taken into account, which can avoid repetitive entropy calculations to obtain a global optimal solution efficiently. To further improve the efficiency, we extend GRRO to filter out inessential labels and features, thus facilitating fast MLFS. We call the extension as GRROfast, in which the key insights are twofold: 1) promising labels and related relevant features are investigated to reduce ineffective calculations in terms of features, even labels and 2) the framework of GRRO is reconstructed to generate the optimal result with an ensemble. Moreover, our proposed algorithms have an excellent mechanism for exploiting the inherent properties of multilabel data; specifically, we provide a formulation to enhance the proposal with label-specific features. Extensive experiments clearly reveal the effectiveness and efficiency of our proposed algorithms.

Graph Convolutional Network Discrete Hashing for Cross-Modal Retrieval.

With the rapid development of deep neural networks, cross-modal hashing has made great progress. However, the information of different types of data is asymmetrical, that is to say, if the resolution of an image is high enough, it can reproduce almost 100% of the real-world scenes. However, text usually carries personal emotion and it is not objective enough, so we generally think that the information of image will be much richer than text. Although most of the existing methods unify the semantic feature extraction and hash function learning modules for end-to-end learning, they ignore this issue and do not use information-rich modalities to support information-poor modalities, leading to suboptimal results, although they unify the semantic feature extraction and hash function learning modules for end-to-end learning. Furthermore, previous methods learn hash functions in a relaxed way that causes nontrivial quantization losses. To address these issues, we propose a new method called graph convolutional network (GCN) discrete hashing. This method uses a GCN to bridge the information gap between different types of data. The GCN can represent each label as word embedding, with the embedding regarded as a set of interdependent object classifiers. From these classifiers, we can obtain predicted labels to enhance feature representations across modalities. In addition, we use an efficient discrete optimization strategy to learn the discrete binary codes without relaxation. Extensive experiments conducted on three commonly used datasets demonstrate that our proposed method graph convolutional network-based discrete hashing (GCDH) outperforms the current state-of-the-art cross-modal hashing methods.

Comparative analysis of the effects of different purification methods on the yield and purity of cow milk extracellular vesicles

AbstractIsolation of extracellular vesicles (EV) has been developing rapidly in parallel with the interest in EVs. However, commonly utilized protocols may not suit more challenging sample matrixes and could potentially yield suboptimal results. Knowing and assessing the pitfalls of isolation procedure to be used, should be involved to some extent for EV analytics. EVs in cow milk are of great interest due to their abundancy and large‐scale availability as well as their cross‐species bioavailability and possible use as drug carriers. However, the characteristics of milk EVs overlap with those of other milk components. This makes it difficult to isolate and study EVs individually. There exists also a lack of consensus for isolation methods. In this study, we demonstrated the differences between various differential centrifugation‐based approaches for isolation of large quantities of EVs from cow milk. Samples were further purified with gradient centrifugation and size exclusion chromatography (SEC) and differences were analyzed. Quality measurements were conducted on multiple independent platforms. Particle analysis, electron microscopy and RNA analysis were used, to comprehensively characterize the isolated samples and to identify the limitations and possible sources of contamination in the EV isolation protocols. Vesicle concentration to protein ratio and RNA to protein ratios were observed to increase as samples were purified, suggesting co‐isolation with major milk proteins in direct differential centrifugation protocols. We demonstrated a novel size assessment of vesicles using a particle mobility analyzer that matched the sizing using electron microscopy in contrast to commonly utilized nanoparticle tracking analysis. Based on the standards of the International Society for Extracellular Vesicles and the quick checklist of EV‐Track.org for EV isolation, we emphasize the need for complete characterization and validation of the isolation protocol with all EV‐related work to ensure the accuracy of results and allow further analytics and experiments.

Progressive Age-Associated Blood-Brain Barrier Leak/Dysfunction-Nexus of Neurodegenerative Disease Using MRI Markers to Identify Preclinical Disease and Potential New Targets for Future Treatments.

This review article focuses on the upstream pertinent pathophysiology leading to neurodegenerative disease. Specifically, the nexus appears to be blood-brain barrier (BBB) leakiness resulting in a two-prong inflammatory disease spectrum damaging the microvasculature and corrupting protein synthesis and degradation with accumulating misfolded toxic proteins. The suboptimal results of removing misfolded proteins mean a new approach to disease in the preclinical state is required aimed at other targets. Validated noninvasive imaging and serologic biomarkers of early preclinical disease implemented in the high-risk patient cohort along with periodic surveillance once effective treatments are developed will be required. This review discusses the physiology and pathophysiology of the BBB, new MRI imaging techniques identifying the leak, and altered fluid dynamic effects in the preclinical state. The risk factors for disease development, preventative measures, and potential treatment targets are also discussed.

The Balloon-Assisted Double-Kissing T-Stenting Technique: Concept, In Vitro Model, and Case Examples

BackgroundIn complex bifurcation percutaneous coronary intervention, 2-stent strategies are often required. Commonly used 2-stent techniques can lead to suboptimal results due to their complexity. We developed the balloon-assisted double-kissing T-stenting (DKT) technique, which uses balloons to optimize stent placement, delivery, and final architecture. MethodsWith the balloon-assisted DKT technique, a balloon is inflated into the main branch (MB) to identify the best position of the side-branch (SB) stent instead of relying on angiography. DKT aims at supporting the SB ostium with stent crowns instead of distorted open cell(s), by achieving a longitudinal deformation with minimal crush of the SB stent upon implantation of the MB stent. This hypothesis was tested on a bench model. We report how the technique was performed in 2 cases and provide intracoronary imaging of the results. ResultsAs hypothesized, DKT resulted in a longitudinal accordion-like deformation and minimal crush effect on bench. The SB ostium was supported by stent crowns. The SB wall opposed to the carina was well covered with crowns from the MB stent after proximal optimization technique and final kissing. The technique was successfully used in 2 complex left main cases with perfect coverage of the SB ostium as assessed with intracoronary imaging. ConclusionsThe balloon-assisted DKT is a simple technique that combines strengths of double-kissing crush and culotte techniques, results in appropriate SB ostium coverage, and deserves further investigation.

Intercomparison of initialization methods for seasonal-to-decadal climate predictions with the NorCPM

Initialization is essential for accurate seasonal-to-decadal (S2D) climate predictions. The initialization schemes used differ on the component initialized, the Data Assimilation method, or the technique. We compare five popular schemes within NorCPM following the same experimental protocol: reanalysis from 1980 to 2010 and seasonal and decadal predictions initialized from the reanalysis. We compare atmospheric initialization—Newtonian relaxation (nudging)—against ocean initialization—Ensemble Kalman Filter—(ODA). On the atmosphere, we explore the benefit of full-field (NudF-UVT) or anomaly (NudA-UVT) nudging of horizontal winds and temperature (U, V, and T) observations. The scheme NudA-UV nudges horizontal winds to disentangle the role of wind-driven variability. The ODA+NudA-UV scheme is a first attempt at joint initialization of ocean and atmospheric components in NorCPM. During the reanalysis, atmospheric nudging improves the synchronization of the atmosphere and land components with the observed data. Conversely, ODA is more effective at synchronizing the ocean component with observations. The atmospheric nudging schemes are better at reproducing specific events, such as the rapid North Atlantic subpolar gyre shift. An abrupt climatological change using the NudA-UV scheme demonstrates that energy conservation is crucial when only assimilating winds. ODA outperforms atmospheric-initialized versions for S2D global predictions, while atmospheric nudging is preferable for accurately initializing phenomena in specific regions, with the technique’s benefit depending on the prediction’s temporal scale. For instance, atmospheric full-field initialization benefits the tropical Atlantic Niño at 1-month lead time, and atmospheric anomaly initialization benefits longer lead times, reducing hindcast drift. Combining atmosphere and ocean initialization yields sub-optimal results, as sustaining the ensemble’s reliability—required for ODA’s performance—is challenging with atmospheric nudging.

BSDSNet: Dual-Stream Feature Extraction Network Based on Segment Anything Model for Synthetic Aperture Radar Land Cover Classification

Land cover classification using high-resolution Polarimetric Synthetic Aperture Radar (PolSAR) images obtained from satellites is a challenging task. While deep learning algorithms have been extensively studied for PolSAR image land cover classification, the performance is severely constrained due to the scarcity of labeled PolSAR samples and the limited domain acceptance of models. Recently, the emergence of the Segment Anything Model (SAM) based on the vision transformer (VIT) model has brought about a revolution in the study of specific downstream tasks in computer vision. Benefiting from its millions of parameters and extensive training datasets, SAM demonstrates powerful capabilities in extracting semantic information and generalization. To this end, we propose a dual-stream feature extraction network based on SAM, i.e., BSDSNet. We change the image encoder part of SAM to a dual stream, where the ConvNext image encoder is utilized to extract local information and the VIT image encoder is used to extract global information. BSDSNet achieves an in-depth exploration of semantic and spatial information in PolSAR images. Additionally, to facilitate a fine-grained amalgamation of information, the SA-Gate module is employed to integrate local–global information. Compared to previous deep learning models, BSDSNet’s impressive ability to represent features is akin to a versatile receptive field, making it well suited for classifying PolSAR images across various resolutions. Comprehensive evaluations indicate that BSDSNet achieves excellent results in qualitative and quantitative evaluation when performing classification tasks on the AIR-PolSAR-Seg dataset and the WHU-OPT-SAR dataset. Compared to the suboptimal results, our method improves the Kappa metric by 3.68% and 0.44% on the AIR-PolSAR-Seg dataset and the WHU-OPT-SAR dataset, respectively.

Devignet: High-Resolution Vignetting Removal via a Dual Aggregated Fusion Transformer with Adaptive Channel Expansion

Vignetting commonly occurs as a degradation in images resulting from factors such as lens design, improper lens hood usage, and limitations in camera sensors. This degradation affects image details, color accuracy, and presents challenges in computational photography. Existing vignetting removal algorithms predominantly rely on ideal physics assumptions and hand-crafted parameters, resulting in the ineffective removal of irregular vignetting and suboptimal results. Moreover, the substantial lack of real-world vignetting datasets hinders the objective and comprehensive evaluation of vignetting removal. To address these challenges, we present VigSet, a pioneering dataset for vignetting removal. VigSet includes 983 pairs of both vignetting and vignetting-free high-resolution (over 4k) real-world images under various conditions. In addition, We introduce DeVigNet, a novel frequency-aware Transformer architecture designed for vignetting removal. Through the Laplacian Pyramid decomposition, we propose the Dual Aggregated Fusion Transformer to handle global features and remove vignetting in the low-frequency domain. Additionally, we propose the Adaptive Channel Expansion Module to enhance details in the high-frequency domain. The experiments demonstrate that the proposed model outperforms existing state-of-the-art methods. The code, models, and dataset are available at https://github.com/CXH-Research/DeVigNet.

Pushing the Limit of Fine-Tuning for Few-Shot Learning: Where Feature Reusing Meets Cross-Scale Attention

Due to the scarcity of training samples, Few-Shot Learning (FSL) poses a significant challenge to capture discriminative object features effectively. The combination of transfer learning and meta-learning has recently been explored by pre-training the backbone features using labeled base data and subsequently fine-tuning the model with target data. However, existing meta-learning methods, which use embedding networks, suffer from scaling limitations when dealing with a few labeled samples, resulting in suboptimal results. Inspired by the latest advances in FSL, we further advance the approach of fine-tuning a pre-trained architecture by a strengthened hierarchical feature representation. The technical contributions of this work include: 1) a hybrid design named Intra-Block Fusion (IBF) to strengthen the extracted features within each convolution block; and 2) a novel Cross-Scale Attention (CSA) module to mitigate the scaling inconsistencies arising from the limited training samples, especially for cross-domain tasks. We conducted comprehensive evaluations on standard benchmarks, including three in-domain tasks (miniImageNet, CIFAR-FS, and FC100), as well as two cross-domain tasks (CDFSL and Meta-Dataset). The results have improved significantly over existing state-of-the-art approaches on all benchmark datasets. In particular, the FSL performance on the in-domain FC100 dataset is more than three points better than the latest PMF (Hu et al. 2022).