Multiple Datasets Research Articles

Secure fuzzy retrieval protocol for multiple datasets

With the diversification of data sources and the massive growth of datasets, data retrieval has become increasingly complex and time-consuming. In the traditional retrieval method, if a user wants to query multiple datasets, the general approach is to retrieve them one by one in order, which may lead to duplication of work and waste of resources. Private set intersection is a specific issue in secure multi-party computation. It allows several participants, each holding different sets, to jointly calculate the intersection of their sets without revealing any information other than the intersection. This method is naturally suitable for data fusion. In this work, we propose a secure fuzzy retrieval protocol for multiple datasets. First, we use private set intersection technology to fuse multiple datasets. Then, we perform secure retrieval based on this fused dataset, effectively avoiding the waste of resources caused by separate retrievals, thereby maximizing resource efficiency. It is worth mentioning that the protocol proposed in this paper can also be used for fuzzy retrieval to improve the user’s search experience. More importantly, the protocol can maximize privacy protection during the retrieval process, including strict protection of sensitive information such as retrieval keywords, ensuring that user data and query intentions will not be leaked during the entire retrieval process. Finally, we provide a rigorous security proof and demonstrate the effectiveness of the protocol through simulation experiments.

Transformer-based abstractive indonesian text summarization

The volume of data created, captured, copied, and consumed worldwide has increased from 2 zettabytes in 2010 to over 97 zettabytes in 2020, with an estimation of 181 zettabytes in 2025. Automatic text summarization (ATS) will ease giving points of information and will increase efficiency at the time consumed to understand the information. Therefore, improving ATS performance in summarizing news articles is the goal of this paper. This work will fine-tune the BART model using IndoSum, Liputan6, and Liputan6 augmented dataset for abstractive summarization. Data augmentation for Liputan6 will be augmented with the ChatGPT method. This work will also use r ecall-oriented understudy of gisting evaluation (ROUGE) as an evaluation metric. The data augmentation with ChatGPT used 10% of the clean news article from the Liputan6 training dataset and ChatGPT generated the abstractive summary based on that input, culminating in over 36 thousand data for the model’s fine-tuning. BART model that was finetuned using Indosum, Liputan6, and augmented Liputan6 dataset has the best ROUGE-2 score, outperforming ORACLE’s model although ORACLE still has the best ROUGE-1 and ROUGE-L score. This concludes that fine-tuning the BART model with multiple datasets will increase the performance of the model to do abstractive summarization tasks.

SBPA: Sybil-Based Backdoor Poisoning Attacks for Distributed Big Data in AIoT-Based Federated Learning System

Federated learning (FL) enables a great deal of distributed independent participants to collaborate in training without sharing data. Malicious adversary can poison the local model by backdoor poisoning attacks and utilize the characteristic that server cannot trace original data to make the poisoned model directly aggregated. Especially in the AIoT-FL network that generates large amounts of data in real-time, such an attack is more powerful. In this paper, we design a sybil-based backdoor poisoning attacks (SBPA) against the above vulnerability. Malicious participants inject backdoor triggers into distributed Big Data to covertly complete data poisoning. During subsequent iterative aggregation, the joint model activates the backdoor during testing to achieve misclassification for the backdoor images. Besides, malicious participants create some sybil nodes to join the aggregation by taking advantage of the vulnerability of system devices to be easily disconnected. They are committed to making the poisoned local models aggregated with higher probability. Their goal works on making the final global model misclassify backdoor images while keeping high classification accuracy on the other non-backdoor samples. We conduct extensive experiments on multiple datasets and exhibit more robust performance than the state-of-the-art in various metrics under both data distribution including i.i.d. and non-i.i.d. scenarios.

Instance Consistency Regularization for Semi-Supervised 3D Instance Segmentation.

Large-scale datasets with point-wise semantic and instance labels are crucial to 3D instance segmentation but also expensive. To leverage unlabeled data, previous semi-supervised 3D instance segmentation approaches have explored self-training frameworks, which rely on high-quality pseudo labels for consistency regularization. They intuitively utilize both instance and semantic pseudo labels in a joint learning manner. However, semantic pseudo labels contain numerous noise derived from the imbalanced category distribution and natural confusion of similar but distinct categories, which leads to severe collapses in self-training. Motivated by the observation that 3D instances are non-overlapping and spatially separable, we ask whether we can solely rely on instance consistency regularization for improved semi-supervised segmentation. To this end, we propose a novel self-training network InsTeacher3D to explore and exploit pure instance knowledge from unlabeled data. We first build a parallel base 3D instance segmentation model DKNet, which distinguishes each instance from the others via discriminative instance kernels without reliance on semantic segmentation. Based on DKNet, we further design a novel instance consistency regularization framework to generate and leverage high-quality instance pseudo labels. Experimental results on multiple large-scale datasets show that the InsTeacher3D significantly outperforms prior state-of-the-art semi-supervised approaches.

DsRNAPredictor-II: An improved predictor of identifying dsRNA and its silencing efficiency for Tribolium castaneum based on sequence length distribution

RNA interference (RNAi) has been widely utilized to investigate gene functions and has significant potential for control of pest insects. However, recent studies have revealed that the target insect species, dsRNA molecule length, target genes, and other experimental factors can affect the efficiency of RNAi mediated control, restricting the further development and application of this technology. Therefore, the aim of this study was to establish a deep learning model using bioinformatics to help researchers identify dsRNA fragments with the highest RNAi efficiency. In this study, we optimized an existing model, namely, dsRNAPredictor, by designing sub-models based on different sequence lengths. Accordingly, the data were divided into two groups: 130–399 bp and 400–616 bp long sequences. Then, one-hot encoding was employed to extract sequence information. The convolutional neural network framework comprising three convolutional layers, three average pooling layers, a flattened layer, and three dense layers was employed as the classifier. By adjusting the parameters, we established two sub-models for different sequence distributions. Using multiple independent test datasets and conducting hypothesis testing, we demonstrated that our model exhibits superior performance and strong robustness to dsRNAPredictor, respectively. Therefore, our model may help design dsRNAs with pre-screening potential and facilitate further research and applications.

Bridging Global Context Interactions for High-Fidelity Pluralistic Image Completion.

We introduce PICFormer, a novel framework for Pluralistic Image Completion using a transFormer based architecture, that achieves both high quality and diversity at a much faster inference speed. Our key contribution is to introduce a code-shared codebook learning using a restrictive CNN on small and non-overlapping receptive fields (RFs) for the local visible token representation. This results in a compact yet expressive discrete representation, facilitating efficient modeling of global visible context relations by the transformer. Unlike the prevailing autoregressive approaches, we proposed to sample all tokens simultaneously, leading to more than 100× faster inference speed. To enhance appearance consistency between visible and generated regions, we further propose a novel attention-aware layer (AAL), designed to better exploit distantly related high-frequency features. Through extensive experiments, we demonstrate that the PICFormer efficiently learns semantically-rich discrete codes, resulting in significantly improved image quality. Moreover, our diverse image completion framework surpasses State-of-the-Art methods on multiple image completion datasets.

RNA-binding protein YBX3 promotes PPARγ-SLC3A2 mediated BCAA metabolism fueling brown adipogenesis and thermogenesis.

Activating brown adipose tissue (BAT) thermogenesis is a promising approach to combat obesity and metabolic disorders. The post-transcriptional regulation of BAT thermogenesis mediated by RNA-binding proteins (RBPs) is still not fully understood. This study explores the physiological role of novel RBPs in BAT differentiation and thermogenesis. We used multiple public datasets to screen out novel RBPs responsible for BAT differentiation and thermogenesis. Invitro loss- and gain-of-function experiments were performed in both C3H10T1/2 preadipocytes and mature brown adipocytes to determine the role of Y-box binding protein 3 (YBX3) in brown adipocyte differentiation and thermogenesis. Adeno-associated virus (AAV)-mediated BAT-specific knockdown or overexpression of Ybx3 was applied to investigate the function of YBX3 invivo. YBX3 is a brown adipocyte-enriched RBP induced by cold stimulation and β-adrenergic signaling. Both invitro loss- and gain-of-function experiments demonstrate that YBX3 is essential for brown adipocyte differentiation and thermogenesis. BAT-specific loss of Ybx3 dampens thermogenesis and exacerbates diet-induced obesity in mice, while overexpression of Ybx3 promotes thermogenesis and confers protection against diet-induced metabolic dysfunction. Transcriptome analysis and mitochondrial stress test indicate that Ybx3 deficiency compromises the mitochondrial oxidative phosphorylation, leading to thermogenic failure. Mechanistically, YBX3 stabilizes the mRNA of Slc3a2 and Pparg, which facilitates branched-chain amino acid (BCAA) influx and catabolism and fuels brown adipocyte differentiation and thermogenesis. YBX3 facilitates BAT fueling BCAA to boost thermogenesis and energy expenditure, which protects against obesity and metabolic dysfunction. Thus, YBX3 could be a promising therapeutic target for obesity.

MSSTGNN: Multi-Scaled Spatio-Temporal Graph Neural Networks for short- and long-term traffic prediction

Accurate traffic prediction plays a crucial role in ensuring traffic safety and minimizing property damage. The utilization of STGNNs in traffic prediction has gained significant attention from researchers aiming to capture the intricate time-varying relationships within traffic data. While existing STGNNs commonly rely on Euclidean distance to assess the similarity between nodes, which may fall short in reflecting POI or regional functions. The traffic network exhibits static from a macro perspective, whereas undergoes dynamic changes in the micro perspective. Previous researchers incorporating self-attention to capture time-varying features for constructing dynamic graphs have faced challenges in overlooking the connections between nodes due to the Softmax polarization effect, which tends to amplify extreme value differences, and fails to accurately represent the true relationships between nodes. To solve this problem, we introduce the Multi-Scaled Spatio-Temporal Graph Neural Networks (MSSTGNN), which aims to comprehensively capture characteristics within traffic from multiscale viewpoints to construct multi-perspective graphs. We employ a trainable matrix to enhance the predefined adjacency matrices and to construct an optimal dynamic graph based on both trend and period. Additionally, a graph aggregate technique is proposed to effectively merge trend and periodic dynamic graphs. TCN is developed to model nonstationary traffic data, and we leverage the skip and residual connections to increase the model depth. A two-stage learning approach and a novel MSELoss function is designed to enhance the model's performance. The experimental results demonstrate that the MSSTGNN model outperforms the existing methods, achieving state-of-the-art performances across multiple real-world datasets.

Learning at a Glance: Towards Interpretable Data-Limited Continual Semantic Segmentation via Semantic-Invariance Modelling.

Continual semantic segmentation (CSS) based on incremental learning (IL) is a great endeavour in developing human-like segmentation models. However, current CSS approaches encounter challenges in the trade-off between preserving old knowledge and learning new ones, where they still need large-scale annotated data for incremental training and lack interpretability. In this paper, we present Learning at a Glance (LAG), an efficient, robust, human-like and interpretable approach for CSS. Specifically, LAG is a simple and model-agnostic architecture, yet it achieves competitive CSS efficiency with limited incremental data. Inspired by human-like recognition patterns, we propose a semantic-invariance modelling approach via semantic features decoupling that simultaneously reconciles solid knowledge inheritance and new-term learning. Concretely, the proposed decoupling manner includes two ways, i.e., channel-wise decoupling and spatial-level neuron-relevant semantic consistency. Our approach preserves semantic-invariant knowledge as solid prototypes to alleviate catastrophic forgetting, while also constraining sample-specific contents through an asymmetric contrastive learning method to enhance model robustness during IL steps. Experimental results in multiple datasets validate the effectiveness of the proposed method. Furthermore, we introduce a novel CSS protocol that better reflects realistic data-limited CSS settings, and LAG achieves superior performance under multiple data-limited conditions.

Convolutional Neural Networks for Sentiment Analysis

Dedicated Identifying aimed at discerning subjective information by analyzing the polarity of opinions conveyed in text. Traditionally, recurrent neural networks (RNNs) have been the dominant approach for this type of analysis because of their ability to process sequential data. However, sentiment analysis has experienced a significant transformation with the introduction of convolutional neural networks (CNNs). Originally developed for image analysis, processing text due to their efficient mechanisms for extracting local features.. This paper explores the role of CNNs in sentiment analysis, evaluating their architecture, methodology, and comparative effectiveness against RNN-based models. We propose a comprehensive CNN-based model for sentiment analysis and examine its potential for sentiment classification tasks across multiple datasets.

Atmospheric new particle formation identifier using longitudinal global particle number size distribution data.

Atmospheric new particle formation (NPF) is a naturally occurring phenomenon, during which high concentrations of sub-10 nm particles are created through gas to particle conversion. The NPF is observed in multiple environments around the world. Although it has observable influence onto annual total and ultrafine particle number concentrations (PNC and UFP, respectively), only limited epidemiological studies have investigated whether these particles are associated with adverse health effects. One plausible reason for this limitation may be related to the absence of NPF identifiers available in UFP and PNC data sets. Until recently, the regional NPF events were usually identified manually from particle number size distribution contour plots. Identification of NPF across multi-annual and multiple station data sets remained a tedious task. In this work, we introduce a regional NPF identifier, created using an automated, machine learning based algorithm. The regional NPF event tag was created for 65 measurement sites globally, covering the period from 1996 to 2023. The discussed data set can be used in future studies related to regional NPF.

Comprehensive assessment of cellular senescence in intestinal immunity and biologic therapy response in ulcerative colitis.

Ulcerative Colitis (UC) is an inflammatory disorder characterized by chronic intestinal inflammation and immune dysregulation. Despite a clear association between cellular senescence and chronic inflammation and immune dysregulation, the mechanisms underlying cellular senescence in UC remain unclear. We screened differentially expressed genes (DEGs) associated with cellular senescence in multiple UC datasets, performed immune infiltration analysis, and constructed clinical diagnostic models. Additionally, we investigated the relationship between key genes related to cellular senescence and disease remission in UC patients undergoing biologic therapy, validating their expression in a single-cell dataset. We identified six DEGs associated with cellular senescence (TWIST1, IGFBP5, MME, IFNG, ME1, FOS). Immune infiltration results indicated strong correlations of four of these genes with immune cells and pathways. Through WGCNA, GO, and KEGG analyses, we found that gene modules strongly associated with the expression of hub genes in cellular senescence were enriched in inflammation-related pathways. In the single-cell dataset, the expression of these six key genes exhibited similarities with Immune infiltration results. Additionally, we constructed a nomogram using these six genes for diagnosing UC, demonstrating good diagnostic capability and clinical efficacy. Kaplan-Meier survival analysis revealed a significant association between changes in the expression levels of these cell genes and disease remission in UC patients undergoing biologic therapy. This study utilizes bioinformatic analysis and machine learning to identify and analyze features associated with cellular senescence in multiple UC datasets. It provides insights into the role of cellular senescence in the premature onset of intestinal aging in UC and offers new perspectives for exploring novel therapeutic targets.

Optimizing Multi-Omics Data Imputation with NMF and GAN Synergy.

Integrating multiple omics datasets can significantly advance our understanding of disease mechanisms, physiology, and treatment responses. However, a major challenge in multi-omics studies is the disparity in sample sizes across different datasets, which can introduce bias and reduce statistical power. To address this issue, we propose a novel framework, OmicsNMF, designed to impute missing omics data and enhance disease phenotype prediction. OmicsNMF integrates Generative Adversarial Networks (GANs) with Non-Negative Matrix Factorization (NMF). NMF is a well-established method for uncovering underlying patterns in omics data, while GANs enhance the imputation process by generating realistic data samples. This synergy aims to more effectively address sample size disparity, thereby improving data integration and prediction accuracy. For evaluation, we focused on predicting breast cancer subtypes using the imputed data generated by our proposed framework, OmicsNMF. Our results indicate that OmicsNMF consistently outperforms baseline methods. We further assessed the quality of the imputed data through survival analysis, revealing that the imputed omics profiles provide significant prognostic power for both overall survival and disease-free status. Overall, OmicsNMF effectively leverages GANs and NMF to impute missing samples while preserving key biological features. This approach shows potential for advancing precision oncology by improving data integration and analysis. Source code is available at: https://github.com/compbiolabucf/OmicsNMF.

LIDeepDet: Deepfake Detection via Image Decomposition and Advanced Lighting Information Analysis

The proliferation of AI-generated content (AIGC) has empowered non-experts to create highly realistic Deepfake images and videos using user-friendly software, posing significant challenges to the legal system, particularly in criminal investigations, court proceedings, and accident analyses. The absence of reliable Deepfake verification methods threatens the integrity of legal processes. In response, researchers have explored deep forgery detection, proposing various forensic techniques. However, the swift evolution of deep forgery creation and the limited generalizability of current detection methods impede practical application. We introduce a new deep forgery detection method that utilizes image decomposition and lighting inconsistency. By exploiting inherent discrepancies in imaging environments between genuine and fabricated images, this method extracts robust lighting cues and mitigates disturbances from environmental factors, revealing deeper-level alterations. A crucial element is the lighting information feature extractor, designed according to color constancy principles, to identify inconsistencies in lighting conditions. To address lighting variations, we employ a face material feature extractor using Pattern of Local Gravitational Force (PLGF), which selectively processes image patterns with defined convolutional masks to isolate and focus on reflectance coefficients, rich in textural details essential for forgery detection. Utilizing the Lambertian lighting model, we generate lighting direction vectors across frames to provide temporal context for detection. This framework processes RGB images, face reflectance maps, lighting features, and lighting direction vectors as multi-channel inputs, applying a cross-attention mechanism at the feature level to enhance detection accuracy and adaptability. Experimental results show that our proposed method performs exceptionally well and is widely applicable across multiple datasets, underscoring its importance in advancing deep forgery detection.

Comparison of cell type and disease subset chromatin modifications in SLE.

Systemic lupus erythematosus (SLE) is an autoimmune disease with protean manifestations. There is little understanding of why some organs are specifically impacted in patients and the mechanisms of disease persistence remain unclear. While much work has been done characterizing the DNA methylation status in SLE, there is less information on histone modifications, a more dynamic epigenetic feature. This study identifies two histone marks of activation and the binding of p300 genome-wide in three cell types and three clinical subsets to better understand cell-specific effects and differences across clinical subsets. We examined 20 patients with SLE and 8 controls and found that individual chromatin marks varied considerably across T cells, B cells, and monocytes. When pathways were examined, there was far more concordance with conservation of TNF, IL-2/STAT5, and KRAS pathways across multiple cell types and ChIP data sets. Patients with cutaneous lupus and lupus nephritis generally had less dramatically altered chromatin than the general SLE group. Signals also demonstrated significant overlap with GWAS signals in a manner that did not implicate one cell type more than the others. The pathways identified by altered histone modifications and p300 binding are pathways known to be important from RNA expression studies and recognized pathogenic mechanisms of disease. NFκB and classical inflammatory pathways were strongly associated with increased peak heights across all cell types but were the highest-ranking pathway for all three antibodies in monocytes according to fgsea analysis. IL-6 Jak/STAT3 signaling was the most significant pathway association in T cells marked by H3K27ac change. Therefore, each cell type experiences the disease process distinctly although in all cases there was a strong theme of classical inflammatory pathways. Importantly, this NFκB pathway, so strongly implicated in the patients with generalized SLE, was much less impacted in monocytes when cutaneous lupus was compared to the general SLE cohort and also less impacted in lupus nephritis compared to general SLE. These studies define important cell type differences and emphasize the breadth of the inflammatory effects in SLE.

PICZL : Image-based photometric redshifts for AGN

Computing reliable photometric redshifts (photo-z) for active galactic nuclei (AGN) is a challenging task, primarily due to the complex interplay between the unresolved relative emissions associated with the supermassive black hole and its host galaxy. Spectral energy distribution (SED) fitting methods, while effective for galaxies and AGN in pencil-beam surveys, face limitations in wide or all-sky surveys with fewer bands available, lacking the ability to accurately capture the AGN contribution to the SED, hindering reliable redshift estimation. This limitation is affecting the many tens of millions of AGN detected in existing datasets, such as those AGN clearly singled out and identified by SRG/eROSITA. Our goal is to enhance photometric redshift performance for AGN in all-sky surveys while simultaneously simplifying the approach by avoiding the need to merge multiple data sets. Instead, we employ readily available data products from the 10th Data Release of the Imaging Legacy Survey for the Dark Energy Spectroscopic Instrument, which covers > 20,000 deg of extragalactic sky with deep imaging and catalog-based photometry in the grizW1-W4 bands. We fully utilize the spatial flux distribution in the vicinity of each source to produce reliable photo-z. We introduce PICZL a machine-learning algorithm leveraging an ensemble of convolutional neural networks. Utilizing a cross-channel approach, the algorithm integrates distinct SED features from images with those obtained from catalog-level data. Full probability distributions are achieved via the integration of Gaussian mixture models. On a validation sample of 8098 AGN PICZL achieves an accuracy $ NMAD $ of 4.5<!PCT!> with an outlier fraction eta of 5.6<!PCT!>. These results significantly outperform previous attempts to compute accurate photo-z for AGN using machine learning. We highlight that the model's performance depends on many variables, predominantly the depth of the data and associated photometric error. A thorough evaluation of these dependencies is presented in the paper. Our streamlined methodology maintains consistent performance across the entire survey area, when accounting for differing data quality. The same approach can be adopted for future deep photometric surveys such as LSST and Euclid, showcasing its potential for wide-scale realization. With this paper, we release updated photo-z (including errors) for the XMM-SERVS W-CDF-S, ELAIS-S1 and LSS fields.

Deep learning-based models for preimplantation mouse and human embryos based on single-cell RNA sequencing.

The rapid growth of single-cell transcriptomic technology has produced an increasing number of datasets for both embryonic development and in vitro pluripotent stem cell-derived models. This avalanche of data surrounding pluripotency and the process of lineage specification has meant it has become increasingly difficult to define specific cell types or states in vivo, and compare these with in vitro differentiation. Here we utilize a set of deep learning tools to integrate and classify multiple datasets. This allows the definition of both mouse and human embryo cell types, lineages and states, thereby maximizing the information one can garner from these precious experimental resources. Our approaches are built on recent initiatives for large-scale human organ atlases, but here we focus on material that is difficult to obtain and process, spanning early mouse and human development. Using publicly available data for these stages, we test different deep learning approaches and develop a model to classify cell types in an unbiased fashion at the same time as defining the set of genes used by the model to identify lineages, cell types and states. We used our models trained on in vivo development to classify pluripotent stem cell models for both mouse and human development, showcasing the importance of this resource as a dynamic reference for early embryogenesis.

Integrated analysis of single-cell and bulk transcriptome reveals hypoxia-induced immunosuppressive microenvironment to predict immunotherapy response in high-grade serous ovarian cancer

BackgroundHypoxia is significantly associated with cancer progression and treatment outcomes. Nevertheless, the precise molecular mechanisms underlying the hypoxia-induced immunosuppressive microenvironment in high-grade serous ovarian cancer (HGSOC) are still not fully understood.MethodsBy analyzing five independent transcriptomic datasets, we investigated the effect of hypoxia on prognosis and tumor microenvironment (TME) in HGSOC. The hypoxia levels and the intercellular communication signaling pathways were studied by using single-cell analysis. Furthermore, the Hypoxia-TME classifier was developed and then validated in the multiple HGSOC datasets. In addition, we also investigated the prognostic significance, genetic variations, signaling pathways, and the potential for immunotherapy benefits in different Hypoxia-TME subgroups.ResultsHypoxia was identified as a crucial risk factor in HGSOC, and strongly correlated with an immunosuppressive microenvironment characterized by alterations in the composition and distribution of immune cells. Single-cell analysis elucidated the heterogeneity inherent within the TME in HGSOC, and demonstrated an association between the hypoxic TME and fibroblasts as well as macrophages. CellChat analysis identified SPP1-CD44 and CXCL12-CXCR4 as the principal signaling axes through which macrophages and fibroblasts interact with T cells, respectively. Moreover, a personalized Hypoxia-TME classifier was constructed and validated through the integration of the hypoxia (18 genes) and TME (7 immune cells) scores. It was observed that patients in the Hypoxialow/TMEhigh subgroup displayed a significantly better prognosis than other subgroups. Different subgroups exhibited unique genomic alterations and variations in signaling pathway differences, including TGF-β and Wnt/β-catenin pathways, which are closely associated with various biological functions. Finally, our results indicated that patients in the Hypoxialow/TMEhigh subgroup exhibit a better response to immunotherapy, suggesting the potential utility of the Hypoxia-TME classifier as a new biomarker in HGSOC.ConclusionOur study revealed hypoxia-induced immunosuppressive microenvironment, and developed Hypoxia-TME classifier to distinguish the prognosis, immune characteristics, and potential benefits of immunotherapy in HGSOC.

Performance Analysis of ML and DL Models: Impact of Linear and Non-Linear Optimizers on Model Efficiency

Introduction: Emphasizing the effect of linear and non-linear optimizers on the performance of machine learning (ML) and deep learning (DL) models, this work addresses the CIFAR-10 image classification task. The work explores over multiple data sets how optimizers influence model behavior and efficiency. Objectives: With a view on five primary assessment metrics—accuracy, precision, recall, F1-score, and AUC the main purpose is to evaluate on SVM and CNN models the performance of linear optimizers (gradient descent, SGD) and non-linear optimizers (Adam, RMSProp). Methods: The study offers fair and complete comparison by way of a consistent evaluation strategy. Analyzing how different optimizer influences model convergence rates, computational cost, and training stability takes front stage in the experimental setting. Designed on CIFAR-10, optimizers are used on ML and DL models whose performance on all metrics is noted. Results: Results show that non-linear optimizers especially Adam much increase CNN model performance by means of faster convergence, higher classification accuracy, and better model stability. While helpful for simpler models, linear optimizers such SGD show slower convergence and limited adaptation with more complicated data. Conclusions: This study provides direction on selecting appropriate optimizers depending on model complexity, job needs, and computing restrictions coupled with important information on optimizing model training in image classification and other areas.

DEUFormer: High‐precision semantic segmentation for urban remote sensing images

AbstractUrban remote sensing image semantic segmentation has a wide range of applications, such as urban planning, resource exploration, intelligent transportation, and other scenarios. Although UNetFormer performs well by introducing the self‐attention mechanism of Transformer, it still faces challenges arising from relatively low segmentation accuracy and significant edge segmentation errors. To this end, this paper proposes DEUFormer by employing a special weighted sum method to fuse the features of the encoder and the decoder, thus capturing both local details and global context information. Moreover, an Enhanced Feature Refinement Head is designed to finely re‐weight features on the channel dimension and narrow the semantic gap between shallow and deep features, thereby enhancing multi‐scale feature extraction. Additionally, an Edge‐Guided Context Module is introduced to enhance edge areas through effective edge detection, which can improve edge information extraction. Experimental results show that DEUFormer achieves an average Mean Intersection over Union (mIoU) of 53.8% on the LoveDA dataset and 69.1% on the UAVid dataset. Notably, the mIoU of buildings in the LoveDA dataset is 5.0% higher than that of UNetFormer. The proposed model outperforms methods such as UNetFormer on multiple datasets, which demonstrates its effectiveness.