Optimizing performance metrics with new distance and similarity measures using control parameters of linear Diophantine fuzzy sets

SECOA: Serial Exponential-Crayfish Optimization Algorithm with ResNet features for medical image registration.

Unsupervised deformable registration model based on multi-scale pyramid and accurate similarity measurement

Measuring the similarity of time series is a fundamental task in numerous information processing applications. Dynamic Time Warping (DTW) is a widely used method for time series similarity measurement, yet its reliance solely on linear Euclidean distance and neglect of directional information often limits its ability in scenarios where subtle variations and trajectory orientation carry semantic significance. To address these limitations, we propose Angle-distance Penalized Metric DTW (APMDTW), a novel similarity measure method that integrates a nonlinear spatial distance metric with an adaptive angle-distance penalty. Specifically, a piecewise logarithmic transformation is introduced to enhance sensitivity to fine-grained local differences, while a parameterized angle-distance penalty, adaptively modulated by spatial distance, incorporates directional consistency into the cost function. This joint modeling of spatial magnitude and geometric orientation yields a more discriminative and robust time series similarity measurement. Experiments on 128 UCR benchmark datasets show that APMDTW outperforms six baseline algorithms on 114 datasets, and improves similarity measurement accuracy by an average of 56.52% over six state-of-the-art DTW variants.

Abstract The q –gram distance between two strings $$s,s^\prime$$ , introduced by Ukkonen in 1992, is an alignment-free string similarity measure which can be computed in linear time, as opposed to the quadratic time necessary for alignment/edit distance. It is based on the $$L_1$$ -distance, or Manhattan-distance, between the multiplicity vectors of fixed-length substrings (so-called q-grams or k-mers ), and has been successfully applied in diverse bioinformatics settings. In this paper, we introduce the threshold q-gram distance (T q D), a new distance measure which is similar to the q -gram distance but uses reduced information on the multiplicities of the q -grams. The new measure retains the linear time computation of the q -gram distance but requires significantly less space. Storage space and accuracy of the measure can be controlled via a user-defined threshold t , which sets a limit on the maximum value of the integers in the multiplicity vectors. In particular, for $$t=1$$ , the comparison is made only on the basis of the sets of uniquely occurring q -grams on the one hand, and of repeated q -grams, on the other. We tested the new distance measure, using the benchmarking tool AFproject of Zielezinski et al. [Genome Biology, 2019], on several real-life data sets for phylogenetic reconstruction and compared the results with those of other k -mer based distance measures. Our experiments show that the new measure T q D compares well to other non-alignment based measures regarding accuracy, while requiring substantially less memory than the classic q -gram distance.

Abstract Few-shot object detection (FSOD) aims to achieve precise detection with minimal annotated data for novel classes, and significant progress has been made in recent years. Current FSOD mainly adheres to the fine-tuning paradigm. However, limited by insufficient samples for novel classes, two problems still exist: (1) fewer novel class data fail to establish robust feature distributions, leading to severe confusion between foreground and background information in the feature space; (2) the model tends to be biased toward base class representations, which reduces the classifier’s capacity for novel objects. To address these issues, we introduce two approaches to extend the existing model: the multi-perspective decoupling module (MPD) and the fine-grained feature metric network (Fac-Net). The MPD adopts horizontal and vertical pooling along with a self-calibrating function during feature propagation, enabling the features to be highly focused on the foreground object regions. During gradient propagation, the MPD applies different gradient control strategies for different components, allowing the model to achieve rapid fitting even in a few-shot setting. In addition, Fac-Net is a few-shot classifier based on metric learning that calibrates the original scores by yielding additional fine-grained similarity scores. Unlike general few-shot models, it integrates a fine-grained feature enhancement block and a fine-grained feature matching algorithm. The former generates stronger feature representations by accurately assigning feature weights through channel interactions at various granularities. The latter abandons conventional pixel-by-pixel similarity computation with semantic-aware feature alignment, ensuring spatial correspondence of semantically similar pixels for more precise similarity measurement. Extensive experiments on Pascal VOC and MS COCO benchmarks demonstrate that our method significantly outperforms most state-of-the-art methods, highlighting its effectiveness.

ABSTRACT The P ∗ index has been widely regarded as the only index capable of capturing the exposure dimension of segregation. It has often been referred to as the isolation index because it isolates (or, more generally, separates) the distributional pattern of one group from the rest of the groups. In other words, it quantifies the degree of a particular group’s exposure to the same group. Therefore, the P ∗ index may be viewed as an index of one-group similarity and an index of multiple-group dissimilarity and multiple-group diversity may be formulated by combining the group-interactions of one-group similarity. After incorporating useful mathematical properties to take the space- and time-varying nature of compositional differences into account, I propose a metric unified approach to the measurement of similarity, dissimilarity, and diversity in different contextual settings.

Nucleic acid molecules, including ribonucleic acid (RNA) and deoxyribonucleic acid (DNA), are essential for various biological processes and can adopt diverse 3D structures that serve as potential drug targets. The direct targeting of nucleic acid structures by small molecules represents an emerging field in drug design with enormous potential for providing therapeutic options for diseases that are currently not addressed, including genetic diseases and viral infections. In the early days of this promising field, a shortage of reliable structural data presents a bottleneck to the direct adaptation of structure-based methods, making the simpler yet powerful ligand-based approach an attractive alternative for virtual screening. In this study, we thoroughly evaluate and benchmark these methods against the reported binding of small molecules to diverse nucleic acid targets. We also compare these methods with structure-based molecular docking. Our results demonstrate that classification performance is significantly influenced by the applied descriptors, the chosen similarity measure, and the specific nucleic acid target. We have also proposed a consensus method that combines the best-performing algorithms of distinct nature. According to our studies, this approach outperforms all other tested methods, providing a valuable framework for nucleic acid-targeted drug discovery.

Biofeedback-based treadmill training generally involves 10 or more sessions to assess its effectiveness during stroke rehabilitation. Improvements are seen in some patients during the assessment, while others do not progress. Our aim in this study is to determine (i) if signs of progress are evident from the initial training session and (ii) whether quantitative measurements between consecutive training sessions can guide interventions for non-progressing patients. The study analyzes Minimum Foot Clearance (MFC) data from 15 stroke patients during their baseline and second training sessions to predict outcomes in the post-assessment phase. Based on early biofeedback training data, we propose a novel approach using cosine similarity (CS), correlation coefficient (CC) and cross-correlation distance (XCRD) measures to predict post-assessment improvements in stroke patients. We also introduce a new real-time adherence assessment metric (RAAM) metric to quantify improvements in adherence to feedback between consecutive training sessions, enabling more targeted interventions. The proposed approach using CS, CC and XCRD adherence indicators demonstrates 100% accuracy in predicting improvement during post-assessments. The results show that patients with MFC values dissimilar to their baseline while adhering to targeted feedback are more likely to improve. The work also indicates that patients who don’t show significant overall improvement may benefit from extended training periods, suggesting the potential for personalized rehabilitation strategies.

Increasing availability of genomic data demands algorithmic approaches that can efficiently and accurately conduct downstream genomic analyses. These analyses, such as evaluating selection pressures within and across genomes, can reveal developmental and environmental pressures. One such commonly used metric to measure evolutionary pressures is based on the ratio of non-synonymous and synonomous substitution rates,dN/dS. Conventionally, thedN/dSratio is used to infer selection pressures employing alignments to estimate total non-synonymous and synonymous substitution rates along protein-coding genes. However, this process can be time consuming and not scalable for larger datasets. Recently, a fast, approximate similarity measure, FracMinHash containment, was introduced and related to average nucleotide identity. In this work, we show how FracMinHash containment can be used to quickly estimatedN/dSenabling alignment-free estimations at a genomic level.Through simulated and real world experiments, our results indicate that employing FracMinHash containment to estimatedN/dSis scalable, enabling pairwisedN/dSestimations for 85,205 genomes within 5 hours. Furthermore, our approach is comparable to traditionaldN/dSmethods, representing sequences subject to positive and negative selection across various mutation rates. Moreover, we used this model to evaluate signatures of selection between Archaeal and Bacterial genomes, identifying a previously unreported metabolic island betweenMethanobrevibacter sp. RGIG2411 andCandidatus Saccharibacteria bacteriumRGIG2249. We present, FracMinHashdN/dS, a novel alignment-free approach for estimatingdN/dSat a genome level that is accurate and scalable beyond gene-level estimations while demonstrating comparability to conventional alignment-baseddN/dSmethods. Leveraging the alignment-free similarity estimation, FracMinHash containment, pairwisedN/dSestimations are facilitated within milliseconds, making it suitable for large-scale evolutionary analyses across diverse taxa. It supports comparative genomics, evolutionary inference, and functional interpretation across both synthetic, and complex biological datasets.Availability and implementationA version of the implementation is available athttps://github.com/KoslickiLab/dnds-using-fmh.git. The reproduction of figures, data, and analysis can be found athttps://github.com/KoslickiLab/dnds-using-fmh_reproducibles.git.Contactdmk333@psu.eduSupplementary informationSupplementary data are available at PLOS Computational Biology online.Author summaryUnderstanding how evolution shapes genomes helps us learn about the pressures organisms face in their environments. Scientists traditionally measure this by comparing genetic changes that alter proteins versus those that don’t, a ratio that reveals whether natural selection is preserving or changing genes. However, this conventional approach requires computationally intensive sequence alignments, making it impractical for analyzing the massive genomic datasets now available. We developed a faster, alignment-free method to estimate evolutionary pressure across entire genomes. Our approach uses a computational technique called FracMinHash that compresses genomic information while preserving meaningful patterns. We tested our method on both simulated and real-world data, including over 85,000 microbial genomes, completing the analysis in just five hours whereas traditional methods would take days or weeks for the same analysis. The results were comparable to traditional methods and correctly identified genes under different types of selection. Using this approach, we discovered a previously unreported shared genetic region between an archaeal and bacterial species from the goat gut microbiome, suggesting ancient gene transfer between these distant branches of life. Our method makes large-scale evolutionary analysis practical for diverse applications, from tracking microbial strains to understanding adaptation in complex microbial communities, potentially accelerating discoveries in comparative genomics and evolutionary biology.

Introduction: The Rural Revitalization Strategy (RRS) aims to address China's three major rural problems, enhance the vitality of agricultural development, and promote agricultural modernization. Within the RRS framework, “digital agriculture” emerges as a critical application of modern science and technology, involving the informatization and standardization of agricultural development. This progression not only supports agricultural modernization but also influences the broader implementation of the RRS. Evaluating the High-Quality Development Level (HQDL) of digital agriculture from the RRS perspective presents a significant challenge in the field of Multiple-Attribute Group Decision Making (MAGDM). Methods: To address the challenges of HQDL evaluation, this study proposes the Intuitionistic Fuzzy CoCoSo approach, based on the Dice similarity measure (IF-DSM-CoCoSo). This method integrates the CoCoSo technique and Dice Similarity Measure (DSM) within the framework of Intuitionistic Fuzzy Sets (IFSs), which are designed to capture and effectively handle fuzzy information in MAGDM scenarios. By combining these techniques, the IF-DSM-CoCoSo model provides a comprehensive and robust solution for evaluating HQDL in digital agriculture. Results: The proposed IF-DSM-CoCoSo approach was validated through a numerical study focused on the HQDL evaluation of digital agriculture from the perspective of the RRS. The results demonstrate the effectiveness and practicality of the model in addressing MAGDM problems, offering reliable insights and solutions for decision-makers. Discussion: The findings highlight the importance of integrating advanced decision-making techniques, such as the IF-DSM-CoCoSo model, to address the complexities of evaluating HQDL in digital agriculture. By leveraging Intuitionistic Fuzzy Sets, the approach effectively handles uncertainty and provides a robust framework for decision-making. However, further research is needed to refine the model and explore its application across diverse agricultural contexts and decision-- making scenarios. result: Through comparative analysis, we validate the effectiveness and superiority of the IF-DSM-CoCoSo model in handling MAGDM with IFSs. Conclusion: This study contributes to the understanding and practice of HQDL evaluation for digital agriculture within the RRS framework. The proposed IF-DSM-CoCoSo model offers valuable insights and practical guidance for decision-makers, advancing the development and implementation of high-quality digital agriculture and supporting the broader goals of the RRS.

Web phishing, a significant form of cybercrime, seeks to obtain illegal access to users’ sensitive information. Despite the abundance of web phishing detection datasets, a systematic framework for their comparative analytical evaluation is lacking. This paper addresses this gap by proposing a novel graph-based framework for analyzing Web phishing detection datasets through network science. We construct Features’ Similarity Graphs (FSGs) for six major datasets – Tamal, Tan, PhiUSIIL, Hannousse, Vrbančič, and Kumar – using three similarity measures: Pearson Correlation Coefficient, Mutual Information, and Kendall’s Tau, which captures linear, non-linear, and rank-order feature interdependencies, respectively. From each FSG and its giant component, we extract 15 key metrics to characterize network structure and feature relationships. Our analysis reveals distinct dataset characteristics, guiding researchers in selecting appropriate datasets for specific research objectives. For instance, Tamal and Vrbančič are ideal for graph-based approaches due to their high graph density, while Tan excels in visual and spatial similarity-based methods. PhiUSIIL is suited for hybrid approaches, Hannousse for deep learning models, and Kumar for lightweight, resource-efficient applications. This framework provides a systematic comparison of Web phishing datasets and offers practical recommendations for the research community, enabling more adaptive phishing detection systems.

Picture hesitant fuzzy sets (PHFSs) provide a powerful framework for modeling uncertainty by incorporating the degrees of membership, non-membership, neutrality, and refusal. This paper introduces a novel similarity measure, the Weighted Tangent Similarity Measure (ωTSM), designed to improve sensitivity to subtle variations and to capture nonlinear interactions amongPHFS components. The proposed measure is theoretically established and empirically validated through two real-world decision-making problems: medical diagnosis and building material classification. In both cases, ωTSM effectively distinguishes between closely related alternatives and consistently outperforms existing methods. The comparative analysis highlights its robustness,flexibility, and practical value in environments characterized by ambiguity and hesitation. Overall, this study advances fuzzy decision-making models and provides a foundation for future applications in knowledge representation and intelligent systems.

This research presents a new theoretical framework and shows how to use it e ectively to data analysis. First, we introduce a novel method to complex neutrosophic soft sets (CNSS) and ensure the generalize ability of their use by de ning basic operations such as union, intersection, and complement. We go on to construct a complete one-value neutrosophic soft topology, de ning the most important topological properties such as interior, closure, and other related theoretical aspects, to give a strong mathematical foundation. An example is then provided to illustrate this frameworks analytical capabilities. Using a comprehensive visualizing package that includes spline-smoothed functional curves, 3D surface plots, and 2D Heatmaps, we apply our method to address a signal-template alignment problem. The Cotangent Similarity Measure (Cot SM) is used to systematically examine three input signals (S1 S3) and three templates (T1 T3). The ndings are striking and unambiguous: a worldwide maximum indicates a high degree of condence, and the correlation between S2 T1 (CotSM = 06653) is widely recognized as the strongest and most conclusive connection. Subsequent analysis shows that while signal S2 is the most important overall, signals S1 and S3 are helpful for particular targets.

This work introduces a novel hybrid emotion signal/template mapping system that combines the artificial intelligence (AI) validation layer with complex two-valued neutrosophic soft sets (DNSS). Next, by describing a comprehensive DNSS topology, fundamental operations, and properties, we create a rigorous mathematical foundation. Its analytical center measures emotional alignment using a Cotangent Similarity Measure (Cot SM), which shows a discernible hierarchy of connectedness between signal channels (S1 − S4) and emotional foci (T1 − T4). The multi-method visualization displays a hierarchy between cases of absolute dissonance (T2S4: 0.1431) and strong and stable pairs (e.g., T2S3: 0.3776). To guarantee the consistency of the paradigms, a nonlinear verification technique was implemented using an Artificial Neural Network (ANN). Strong performance metrics (Precision = 0.86, Recall = 0.83, F1-Score = 0.84, ROC-AUC = 0.91) verified that ANN was able to replicate the analytical hierarchy. The integration of the symbolic cotangent model with numerical ANN validation demonstrates statistical consistency and structural harmony across computation domains. In order to establish a repeatable paradigm of affective computing and quantification of uncertainty in emotional connections, this research presents a theoretically and AI-checkable model of emotional analytic.

Traditional research has difficulty in precisely evaluating students’ movements in martial arts teaching, and cannot provide targeted teaching suggestions based on students’ characteristics in personalized guidance. This study constructs a movement evaluation and personalized guidance system for martial arts teaching by integrating the EfficientNet model with a recommendation algorithm based on collaborative filtering. First, EfficientNet is used to perform multi-category classification evaluation on martial arts movements, and the cross-entropy loss function is used to improve the classification accuracy. Then, combined with the students’ historical movement performance, a collaborative filtering algorithm based on the Pearson correlation coefficient is used to recommend personalized training plans for each student. Grid search is used to optimize hyperparameters such as user neighborhood size and similarity measurement to ensure the accuracy and diversity of recommendation results. Experimental data shows that after systematic teaching evaluation and personalized guidance, the average score of students increases by about 15.84%, and the average self-assessed satisfaction (1-5) of students after the recommended content reaches 4.1. The system can accurately evaluate students’ movement performance and provide targeted training suggestions based on individual needs.

Relation classification (RC) is a fundamental task in knowledge discovery. Prototypical networks are commonly used for few-shot RC tasks to tackle labeled data scarcity and long-tail relations, but they always overlook classification reliability and outlier features, potentially leading to sub-optimal similarity measurement. In a dynamic world with emerging novel semantic relations, incremental few-shot learning of relations gains attention, involving learning of both base and novel relations in new tasks while retaining base relation knowledge. However, this becomes challenging with increasing novel relations, requiring that models should reduce reliance on learning base relations for the new task. Therefore, this paper explores class-incremental few-shot relation classification (CIFRC) using a Gaussian-augmented prototypical network (GA-Proto). GA-Proto refines the similarity measurement by incorporating discrepancies between ideal and actual classifications via Gaussian Mixture Model and analyzing Gaussian outlier features of query instances. It also employs reliability learning and knowledge distillation to mitigate encoding space distortion and base relation forgetting by enhancing classification reliability and transferring base relation knowledge, respectively. Additionally, GA-Proto uses label smoothing to alleviate novel relation overfitting. Experimental results on three public datasets demonstrate that GA-Proto outperforms existing methods on CIFRC, achieving up to 19.64% improvement in accuracy. The datasets and source code for GA-Proto are released at https://github.com/GTZN2/GA-Proto .

Abstract This study analyzes and evaluates prominent teaching modes to identify the optimal approach for implementation, aiming to enhance pedagogical outcomes. It proposes a systematic framework for teaching evaluation and mode innovation through a multi-criteria decision-making (MCDM) methodology that integrates q-rung orthopair fuzzy sets (q-ROFS) with the Ranking of Alternatives based on Trace to Median Index (RATMI) approach. This integrated method selects the optimal C programming language teaching mode to maximize instructional effectiveness, where RATMI evaluates teaching modes, while q-ROFS reduces uncertainty in expert assessments. In addition, a novel q-ROFS score function derived from the proposed similarity measure is developed to enhance discrimination capability and overcome data limitations. Furthermore, teaching factor weights are computed by combining subjectively assigned weights from teaching experts with objectively derived weights generated through a new entropy measure. Finally, a case study on C programming language teaching modes validates the model’s scientific rigor and operational effectiveness via comprehensive sensitivity and comparative analyses.

The rapid spread of fake news undermines public trust and highlights the need for more reliable detection models. Traditional approaches, such as LSTM with standard cosine similarity using Euclidean distance, often fail to capture subtle textual relationships. This study introduces a Modified Cosine Similarity (MCS) that replaces Euclidean distance with Sørensen-Dice distance and evaluates its effectiveness across four models: Long Short-Term Memory (LSTM), K-Nearest Neighbor (KNN), Random Forest (RF), and Support Vector Machine (SVM). Baseline results using cosine similarity showed strong performance, with SVM achieving the highest accuracy (0.987) and F1-score (0.986), followed by RF (accuracy 0.979) and KNN (accuracy 0.973). However, enhanced models with MCS demonstrated substantial improvements. LSTM achieved the best results overall (accuracy 0.997, recall 0.998, F1-score 0.997) with reduced cross-entropy loss (0.016), false positive rate (0.005), and false negative rate (0.002). SVM and KNN also showed notable gains with accuracies of 0.995 and 0.991, respectively, while RF recorded high recall (0.995) and competitive performance across metrics. These findings confirm that integrating Sørensen-Dice distance into cosine similarity significantly boosts semantic representation and model performance, making MCS a robust similarity measure for advancing fake news detection.

Similarity measures combined with refusal measures of picture fuzzy sets and their applications