JournalTOCs

International Journal of Machine Learning and Cybernetics
Journal Prestige (SJR): 0.7

Citation Impact (citeScore): 2
Number of Followers: 31

Hybrid journal (It can contain Open Access articles)
ISSN (Print) 1868-8071 - ISSN (Online) 1868-808X
Published by Springer-Verlag

[2468 journals]

Adaptive locally connected recurrent unit (ALCRU)
- Free pre-print version: Loading...
  Rate this result: What is this?
  
  Please help us test our new pre-print finding feature by giving the pre-print link a rating.
  A 5 star rating indicates the linked pre-print has the exact same content as the published article.
  
  Abstract: Research has shown that adaptive locally connected neurons outperform their fully connected (dense) counterparts, motivating this study on the development of the Adaptive Locally Connected Recurrent Unit (ALCRU). ALCRU modifies the Simple Recurrent Neuron Model (SimpleRNN) by incorporating spatial coordinate spaces for input and hidden state vectors, facilitating the learning of parametric local receptive fields. These modifications add four trainable parameters per neuron, resulting in a minor increase in computational complexity. ALCRU is implemented using standard frameworks and trained with back-propagation-based optimizers. We evaluate the performance of ALCRU using diverse benchmark datasets, including IMDb for sentiment analysis, AdditionRNN for sequence modelling, and the Weather dataset for time-series forecasting. Results show that ALCRU achieves accuracy and loss metrics comparable to GRU and LSTM while consistently outperforming SimpleRNN. In particular, experiments with longer sequence lengths on AdditionRNN and increased input dimensions on IMDb highlight ALCRU’s superior scalability and efficiency in processing complex data sequences. In terms of computational efficiency, ALCRU demonstrates a considerable speed advantage over gated models like LSTM and GRU, though it is slower than SimpleRNN. These findings suggest that adaptive local connectivity enhances both the accuracy and efficiency of recurrent neural networks, offering a promising alternative to standard architectures.
  PubDate: 2025-07-03
Remaining useful life prediction of rolling bearing with determined first
predicting time by transfer clustering learning and stochastic
configuration networks
- Free pre-print version: Loading...
  Rate this result: What is this?
  
  Please help us test our new pre-print finding feature by giving the pre-print link a rating.
  A 5 star rating indicates the linked pre-print has the exact same content as the published article.
  
  Abstract: In the prediction of the Remaining Useful Life (RUL) of rolling bearings, determining the First Predicting Time (FPT) for bearing life under different working conditions is challenging, leading to low prediction accuracy. To address this issue, this paper proposes a method based on Transfer Fuzzy C-Means-Relative Membership Difference (TFCM-RMD) for classifying health states and selecting appropriate FPTs. The Stochastic Configuration Networks (SCNs) model is employed to predict the RUL of rolling bearings. Firstly, Complete Ensemble Empirical Mode Decomposition with Adaptive Noise (CEEMDAN) is utilized to decompose the original vibration signal of the bearing and extract its features. Secondly, due to the fluctuation of the health index during degradation, there exists a fuzzy clustering boundary problem between different states of rolling bearings. Therefore, the TFCM-RMD method is adopted in this study. By quantifying the difference in membership degrees, the FPT for bearings under various working conditions is selected. Finally, the SCNs prediction model is established using data after the FPT. In this study, the proposed method is experimentally validated using PHM2012 challenge data and compared with other prediction methods. The experimental results demonstrate that the proposed method not only achieves accurate classification of bearing health states under different working conditions but also improves the RUL prediction scores of rolling bearings by 6%, 5%, 18%, 13%, and 10%, respectively, compared to other methods. This approach enables earlier detection of bearing anomalies and holds practical significance for predictive maintenance.
  PubDate: 2025-07-03
SGSA-C: a new spiky gravitational search algorithm clustering method for
segmentation
- Free pre-print version: Loading...
  Rate this result: What is this?
  
  Please help us test our new pre-print finding feature by giving the pre-print link a rating.
  A 5 star rating indicates the linked pre-print has the exact same content as the published article.
  
  Abstract: Automatic clustering into an optimal number of clusters poses a significant challenge. Various metaheuristic-based methods, such as the gravitational search algorithm, differential evolution, firefly algorithm, and particle swarm optimization, have been utilized in the literature to address this challenge. However, these methods suffer from poor solution precision due to the complexity of data, which limits their ability to achieve optimal clustering. To overcome this limitation, a novel approach called spiky gravitational search algorithm-based clustering is introduced. The proposed method employs a new variant of the gravitational search algorithm, referred to as the spiky gravitational search algorithm, to generate optimal clusters. The effectiveness of the proposed method is evaluated on two sets of benchmark functions, namely CEC-2015 and CEC-2019. Furthermore, its clustering performance is tested against six existing clustering methods on five UCI datasets, using three clustering metrics namely, adjusted rand index, normalized mutual information, and clustering accuracy. The proposed method outperforms existing techniques, achieving clustering accuracies of 96.3%, 94.2%, 91.6%, 97.5%, and 94.8% on the five UCI datasets, respectively. In addition, its segmentation performance is assessed on seven images both qualitatively and quantitatively. The proposed method surpasses the considered approaches, reporting the highest average values for Dice coefficient (0.803), structural similarity index measure (0.735), and feature similarity index measure (0.811). Experimental results highlight the robustness, efficiency, and superior performance of the proposed method.
  PubDate: 2025-07-01
A three-way decision model oriented to the twin fuzzy concepts of q-rung
orthopair
- Free pre-print version: Loading...
  Rate this result: What is this?
  
  Please help us test our new pre-print finding feature by giving the pre-print link a rating.
  A 5 star rating indicates the linked pre-print has the exact same content as the published article.
  
  Abstract: In recent years, the three-way decision model has been widely used to address multi-criteria decision-making problems. However, existing models often overlook differences in the minimum requirements and risk aversion of decision-makers (DMs) across different criteria. Moreover, with the increasing complexity and uncertainty of decision problems, the accurate expression of evaluation values has become a critical challenge. Q-rung orthopair fuzzy sets (q-ROFSs), as an extension of intuitionistic fuzzy sets (IFSs) and Pythagorean fuzzy sets (PFSs), offer stronger expressiveness and broader application scenarios. With this in mind, this paper proposes a three-way decision model oriented to the twin fuzzy concepts of q-rung orthopair. Specifically, we first define the twin fuzzy concepts to represent the minimum requirements of DMs and risk aversion coefficients for different criteria, and propose a new method for calculating relative loss functions. Next, based on the TOPSIS semantics, the positive and negative ideal correlation coefficients are constructed. A requirement correlation coefficient, which takes into account the needs of DMs, is also proposed, from which a novel method for calculating the grey conditional probability is developed. Furthermore, we construct three distinct three-way decision models based on three decision perspectives. In addition, the rationality and effectiveness of the proposed model are demonstrated using a supplier selection case, and the practicality of the model is verified using six data sets. The experimental results show that the SRCC values between the ranking results of the proposed method and the comparison methods are mostly greater than 0.7, further demonstrating the effectiveness of the model.
  PubDate: 2025-06-30
A step towards the integration of machine learning and classic model-based
survey methods
- Free pre-print version: Loading...
  Rate this result: What is this?
  
  Please help us test our new pre-print finding feature by giving the pre-print link a rating.
  A 5 star rating indicates the linked pre-print has the exact same content as the published article.
  
  Abstract: The usage of machine learning methods in traditional surveys including official statistics, is still very limited. Therefore, we propose a predictor supported by these algorithms, which can be used to predict any population or subpopulation characteristics. Machine learning methods have already been shown to be very powerful in identifying and modelling complex and nonlinear relationships between the variables, which means they have very good properties in case of strong departures from the classic assumptions. Therefore, we analyse the performance of our proposal under a different set-up, which, in our opinion, is of greater importance in real-life surveys. We study only small departures from the assumed model to show that our proposal is a good alternative, even in comparison with optimal methods under the model. Moreover, we propose the method of the ex ante accuracy estimation of machine learning predictors, giving the possibility of the accuracy comparison with classic methods. The solution to this problem is indicated in the literature as one of the key issues in integrating these approaches. The simulation studies are based on a real, longitudinal dataset, where the prediction of subpopulation characteristics is considered.
  PubDate: 2025-06-30
Predicting risk of coronary heart disease using a teaching–learning
seagull optimization algorithm with the lightboost based gradient boosting
machine
- Free pre-print version: Loading...
  Rate this result: What is this?
  
  Please help us test our new pre-print finding feature by giving the pre-print link a rating.
  A 5 star rating indicates the linked pre-print has the exact same content as the published article.
  
  Abstract: Coronary Heart Disease, a common cause of death globally, has a debilitating effect on the individual, society, and the economy. Hence, screening and treating Coronary Heart Disease is critical. This paper proposes a risk prediction method of Coronary Heart Disease based on a new machine learning model, which combines a teaching–learning Seagull Optimization Algorithm with the Lightboost based Gradient Boosting Machine. The beta distribution in the Seagull Optimization Algorithm initializes the population to maximize the enveloped optimal solution. A variable convergence factor is applied to balance the global search and local search speed by using a logistic function to improve the algorithm’s search performance. A teaching–learning strategy is integrated to enhance the diversity of the seagull population and the quality of the seagulls during the seagull position updates to avoid local optimality. The teaching–learning Seagull Optimization Algorithm is used to search for the best combination of the main hyperparameters of Lightboost based Gradient Boosting Machine model. An empirical analysis is conducted using the Coronary Heart Disease data set on the Kaggle platform. The result shows that the Accuracy, The Positive Rate, True Negative Rate, F1-score, G-mean and Area Under Curve value of the proposed paper are 0.8978, 0.9333, 0.9388, 0.8934, 0.9361 and 95.74% respectively, which performs better than the other compared machine learning models. This validates the feasibility and superiority of the proposed model in risk prediction of Coronary Heart Disease.
  PubDate: 2025-06-29
Multi-scale multi-label feature selection via fuzzy mutual information
- Free pre-print version: Loading...
  Rate this result: What is this?
  
  Please help us test our new pre-print finding feature by giving the pre-print link a rating.
  A 5 star rating indicates the linked pre-print has the exact same content as the published article.
  
  Abstract: Multi-label feature selection has gained significant attention in recent years due to the growing volume of data. Real-world phenomena often manifest intricate patterns and relationships across multiple scales or levels of granularity. Despite this complexity, existing multi-label feature selection algorithms fall short in harnessing multi-scale information effectively. Moreover, prevalent information-theoretic methods necessitate the discretization of continuous features, risking information loss in the process. In light of these challenges, a new multi-scale multi-label feature selection (MsMLFS) algorithm based on fuzzy information measures is presented. First, the multi-scale multi-label decision table and fuzzy information measures associated with it were defined. Then, the axiomatic model for multi-scale multi-label feature selection is established. Utilizing this, a multi-scale multi-label feature selection algorithm that prioritizes maximum dependency and minimal redundancy is presented. The proposed algorithm is evaluated against nine existing methods on ten benchmark multi-label datasets from various domains. The experimental findings clearly demonstrate the effectiveness of the proposed method on different evaluation metrics. Additionally, the statistical significance and stability of the proposed method is validated, further affirming its efficacy and reliability.
  PubDate: 2025-06-28
Robust visible-infrared person re-identification via frequency-space joint
disentanglement and fusion network
- Free pre-print version: Loading...
  Rate this result: What is this?
  
  Please help us test our new pre-print finding feature by giving the pre-print link a rating.
  A 5 star rating indicates the linked pre-print has the exact same content as the published article.
  
  Abstract: Visible-Infrared person re-identification holds significant importance in domains like security surveillance and intelligent retrieval. Existing methods mainly focus on utilizing spatial information to mitigate modality discrepancies and extract modality-shared features, overlooking the vital person discriminative information embedded in the frequency domain. Additionally, these methods also lack sufficient robustness, making them prone to the adverse effects of noise and damage. To address this issue, we propose a novel Frequency-Space Joint Disentanglement and Fusion Network (FSDF) to explore the key information in both spatial and frequency domains. Specifcally, we design a Frequency and Spatial Information Fusion (FSIF) module to fuse the crucial identity information contained in the frequency and spatial domain using the Fast Fourier Transform (FFT) and feature fusion. Furthermore, as noise commonly manifests as high-frequency information, we design a High-low Frequency Information Disentanglement Mining (HFIDM) module to disentangle high- and low-frequency information and extract crucial robust features, effectively mitigating modal differences and reducing the impact of noise. Extensive experimental results have shown that the proposed FSDF not only outperforms other state-of-the-art methods on the SYSU-MM01, RegDB, and LLCM datasets but also maintains competitiveness in challenging corrupt scenes.
  PubDate: 2025-06-26
RPGT: a retrospective prompt-guided parameter tuning method for knowledge
transfer of contrastive self-supervised vision model
- Free pre-print version: Loading...
  Rate this result: What is this?
  
  Please help us test our new pre-print finding feature by giving the pre-print link a rating.
  A 5 star rating indicates the linked pre-print has the exact same content as the published article.
  
  Abstract: Recently, self-supervised learning (SSL) vision model has achieved great success, but its knowledge transfer usually needs an expensive parameter tuning cost. It hampers the broad application of SSL pre-training vision models. Some recent works attempt to extend the existing prefix prompt method to perform parameter-efficient tuning. However, these methods usually produce prompts and insert prompts in a model-agnostic way, and lack across-task information propagation pipelines. To address the above issues, we propose a Retrospective Prompt-Guided Parameter Tuning (RPGT) method to excavate the potential of contrastive self-supervised vision models on multiple downstream tasks. Firstly, RPGT dynamically generates retrospective prompt (RP) for contrastive SSL models in a model property-aware manner, and heuristically performs prompt insertion and interaction. It facilitates knowledge retrospection while avoiding the tedious prompt search process. Subsequently, a multi-task information propagation (MT-IP) module is proposed to construct across-task information propagation pipeline. It handles the knowledge sharing across tasks, accelerating the learning process of new tasks. Experimental results on twelve datasets demonstrate the effectiveness and generalization of RPGT. Our RPGT method can facilitate the transfer performance of various contrastive SSL models on multiple scenarios while significantly reducing parameter costs.
  PubDate: 2025-06-25
Immunization of binarized deep neural networks against model replication
attacks based on stochastic magnetoresistive RAMs
- Free pre-print version: Loading...
  Rate this result: What is this?
  
  Please help us test our new pre-print finding feature by giving the pre-print link a rating.
  A 5 star rating indicates the linked pre-print has the exact same content as the published article.
  
  Abstract: With the rapid advancement of deep neural networks (DNNs), security has become a critical concern due to the increasing threat of IP theft and reverse engineering in widely deployed DNNs. This paper presents an efficient method to secure the parameters of DNNs using magnetic tunnel junctions (MTJs), characterized by low power consumption, minimal hardware overhead, and rapid accuracy degradation upon intrusion, efficiently countering reverse engineering attempts. The proposed method utilizes the stochastic behavior of MTJs in the sub-critical current regime to secure binarized neural networks against model replication attacks seeking to reverse engineer the network’s synaptic weights. In this method, when the DNNs are not under attack, the weights are updated normally, and the accuracy of the network does not change. However, when DNNs are attacked and compromised, the proposed control and write circuitry updates the weights stochastically, resulting in a significant decrease in accuracy. On average, the accuracy of the tested DNNs declines by approximately 60% compared to the original network accuracy. Our comprehensive simulations showcase the effectiveness of the proposed method in countering model replication attacks and probing attacks, highlighting its superiority over previous approaches. Our proposed method shows that the attacker cannot recover the neural network by changing less than 1% of the network weights.
  PubDate: 2025-06-25
Robust humanoid vehicle ingress with regression and whole body control
- Free pre-print version: Loading...
  Rate this result: What is this?
  
  Please help us test our new pre-print finding feature by giving the pre-print link a rating.
  A 5 star rating indicates the linked pre-print has the exact same content as the published article.
  
  Abstract: Humanoid vehicle ingress-a prerequisite for autonomous driving-remains a critical unsolved challenge due to its sensitivity to initial conditions and unstructured environments. While model-based controllers fail under pose variations, Deep Reinforcement Learning (DRL) methods face computational bottlenecks and dependency on continuous localization. This paper proposes a hybrid control framework that bridges adaptability and efficiency by integrating whole-body control (WBC) and regression-driven adjustments. Our method trains regression models on DRL-generated simulation data to map initial pose deviations to future movement corrections, eliminating real-time DRL inference and dependency on continuous localization. In simulation, the controller achieves 97.6% ingress success with varying starting poses. Real-world experiments also validate that the proposed method enables humanoid vehicle ingress from different starting poses.
  PubDate: 2025-06-24
Long sequence time-series forecasting of rare earth price based on
variational mode decomposition and improved random forest
- Free pre-print version: Loading...
  Rate this result: What is this?
  
  Please help us test our new pre-print finding feature by giving the pre-print link a rating.
  A 5 star rating indicates the linked pre-print has the exact same content as the published article.
  
  Abstract: Given the complex and prolonged industrial processes involved in rare earth production, including the extraction and separation stages, the utility of short-term price predictions is limited due to the extensive time required to adjust production schedules. Consequently, accurately forecasting the long-term price trends of rare earth products is a pressing challenge. To address this, this paper introduces a VMD–SRF hybrid model tailored for long sequence time-series forecasting (LSTF). To simplify the complexity of the initial data and improve the model’s predictive accuracy, variational mode decomposition (VMD) is first employed to analyze the periodicity and random components in price time series; then, it combines the series random forest model, which is improved based on the random forest (RF) algorithm. Series random forest (SRF) model uses dynamic time warping (DTW) distance as heuristic information to address the deficiencies of random forest in long time series forecasting. This hybrid approach, leveraging the strengths of both VMD and SRF, enhances the handling of LSTF issues. An experimental comparative analysis using four representative datasets of rare earth product prices indicates superior prediction accuracy of the proposed method. These advancements present a promising and applicable strategy for addressing LSTF challenges in various practical settings.
  PubDate: 2025-06-24
Eikd: method of extracting important information from feature map for
knowledge distillation
- Free pre-print version: Loading...
  Rate this result: What is this?
  
  Please help us test our new pre-print finding feature by giving the pre-print link a rating.
  A 5 star rating indicates the linked pre-print has the exact same content as the published article.
  
  Abstract: Previous knowledge distillation research has primarily focused on transferring information from the teacher network’s final layer, often neglecting the valuable guidance embedded within the intermediate layers. This oversight can lead to a lack of direction for the student model in focusing on crucial knowledge aspects. Therefore, this paper investigates methods for effectively extracting guidance information from the teacher network’s intermediate layers and proposes the method of extracting important information from feature map for knowledge distillation (EIKD). EIKD utilizes a rank-based pruning method to extract matching guidance information from the teacher network’s intermediate layer feature map based on the size of the student’s feature map. This approach addresses the challenge of knowledge transfer caused by mismatched feature map sizes between the teacher and student models. Furthermore, to prevent the loss of correlation information between the teacher’s feature map channels during the pruning process, EIKD establishes a correlation information matrix for the extracted feature maps and guides the student to fit the relationships between the teacher’s intermediate feature map channels. This ensures that the student effectively receives both the guidance information and the feature map correlation information from the teacher’s intermediate layers. In image classification tasks, using ResNet-34 as the teacher, our method improves the ImageNet Top-1 accuracy of ResNet18 by an average of 2.14%. Additionally, EIKD achieves an average improvement of 3.16% in student accuracy on the CIFAR-100 experiment, surpassing the classic Knowledge Distillation method by an average of 1.45%.
  PubDate: 2025-06-23
Fuzzy $$\beta$$ covering based self-information for feature selection
- Free pre-print version: Loading...
  Rate this result: What is this?
  
  Please help us test our new pre-print finding feature by giving the pre-print link a rating.
  A 5 star rating indicates the linked pre-print has the exact same content as the published article.
  
  Abstract: In recent years, there has been considerable scholarly interest in the exploration of fuzzy $$\beta$$ covering, which synergizes fuzzy set theory with rough set theory to formulate the concept of a fuzzy $$\beta$$ neighborhood. This paper introduces an innovative discernibility measure pertaining to fuzzy $$\beta$$ coverings, aimed at characterizing the distinguishing capability of a fuzzy $$\beta$$ covering family. To this end, the parameterized fuzzy $$\beta$$ neighborhood is introduced as a methodological tool to delineate the similarity between samples and evaluate the distinguishing capacity of a particular fuzzy $$\beta$$ covering family. Subsequently, we propose a novel uncertainty measure, termed relative decision self-information with respect to fuzzy $$\beta$$ covering, by employing fuzzy rough approximations and the concept of self-information. This measure is designed to assess the classification efficacy of attribute subsets. A comprehensive analysis of the measure is conducted, highlighting its enhanced effectiveness in attribute reduction due to its integration of both lower and upper approximations of a fuzzy decision. Finally, an attribute reduction algorithm is utilized to address the issue of redundant fuzzy coverings. Comprehensive experimental results indicate that the proposed method effectively assesses uncertainty across diverse datasets and demonstrates enhanced efficiency in attribute reduction relative to several existing algorithms.
  PubDate: 2025-06-22
Evaluating zero-shot multilingual aspect-based sentiment analysis with
large language models
- Free pre-print version: Loading...
  Rate this result: What is this?
  
  Please help us test our new pre-print finding feature by giving the pre-print link a rating.
  A 5 star rating indicates the linked pre-print has the exact same content as the published article.
  
  Abstract: Aspect-based sentiment analysis (ABSA), a sequence labeling task, has attracted increasing attention in multilingual contexts. While previous research has focused largely on fine-tuning or training models specifically for ABSA, we evaluate large language models (LLMs) under zero-shot conditions to explore their potential to tackle this challenge with minimal task-specific adaptation. We conduct a comprehensive empirical evaluation of a series of LLMs on multilingual ABSA tasks, investigating various prompting strategies, including vanilla zero-shot, chain-of-thought (CoT), self-improvement, self-debate, and self-consistency, across nine different models. Results indicate that while LLMs show promise in handling multilingual ABSA, they generally fall short of fine-tuned, task-specific models. Notably, simpler zero-shot prompts often outperform more complex strategies, especially in high-resource languages like English. These findings underscore the need for further refinement of LLM-based approaches to effectively address ABSA task across diverse languages.
  PubDate: 2025-06-18
Fourier-based degradation-aware transformer-style network for blind image
super-resolution
- Free pre-print version: Loading...
  Rate this result: What is this?
  
  Please help us test our new pre-print finding feature by giving the pre-print link a rating.
  A 5 star rating indicates the linked pre-print has the exact same content as the published article.
  
  Abstract: In recent years, the advance of convolutional neural networks (CNNs) helped image super-resolution (SR) research to achieve remarkable improvement. However, the majority of the SR methods are non-blind, assuming the image degradation is defined (e.g., bicubic). So, these methods struggle in case of unknown degradation. Recently, a blind SR task was developed to deal with this problem using degradation estimation. Although many models have been developed for blind SR, blind SR is still a challenging problem and needs to be improved further. Therefore, this paper proposes a Fourier-based Degradation-aware Transformer-style Network (FDATSRN) for a blind image SR. The idea of the FDATSRN is based on exploring the spatial context of the input image in the Fourier space and a large receptive field for restoring the SR image. This is achieved by designing a Fourier-based degradation-aware Transformer block (FDATB) to be the backbone of the FDATSRN model. The FDATB is designed to be a lightweight version of the SR-transformer block based on using the degradation-aware convolution, convolutional modulation, and Fourier unit. Extensive experiments are performed to show the efficiency of the proposed FDATSRN in handling a large receptive field.
  PubDate: 2025-06-17
Towards high-utility sequential rules with repetitive items
- Free pre-print version: Loading...
  Rate this result: What is this?
  
  Please help us test our new pre-print finding feature by giving the pre-print link a rating.
  A 5 star rating indicates the linked pre-print has the exact same content as the published article.
  
  Abstract: Discovering sequential rules in the sequence database is of key importance for a variety of fields, ranging from customer behavior analysis to intrusion detection. To obtain more informative rules, high-utility sequential rule mining (HUSRM) was proposed. Its goal is to find those sequential rules with high utility values and high confidence, i.e., HUSRs. As far as we know, there are a few algorithms proposed to discover HUSRs. However, these algorithms do not fully consider the existence of repeated items in the sequences of the database. In this paper, we propose an algorithm named USER to discover HUSRs in multi-sequences with the existence of repeated items. A data structure called an occurrence information (OI)-list is designed to distinguish the different occurrences of items in a sequence. Moreover, the change of the upper bound value after the rule expansion is discussed in detail, which is complicated by the repeated items. We also propose four pruning strategies (ROOR, REIO-I, REIO-II, and LEIO) to optimize mining efficiency when there are too many repeated items in the sequence. Finally, we conduct experiments on several datasets, and the results show that USER can mine HUSRs with more accurate utility values in an acceptable amount of time and memory.
  PubDate: 2025-06-17
Path planning problem solved by an improved black-winged kite optimization
algorithm based on multi-strategy fusion
- Free pre-print version: Loading...
  Rate this result: What is this?
  
  Please help us test our new pre-print finding feature by giving the pre-print link a rating.
  A 5 star rating indicates the linked pre-print has the exact same content as the published article.
  
  Abstract: The Black Kite Algorithm (BKA) is a nature-inspired meta-heuristic algorithm designed to mimic the migratory and predatory behaviours of the black kite. However, the parameter adjustments and heavy dependence on the previous generation’s position update strategy in BKA can negatively affect the stability of optimization results. This paper introduces an improved algorithm named ALBKA to address these limitations. ALBKA enhances population diversity during initialization by employing an improved Tent chaotic map with increased randomness and uniformity. It integrates dynamic adjustments of search amplitudes and adaptively modifies neighbourhood and global guidance factors. Furthermore, a Lévy flight strategy is introduced to help the algorithm escape local optima, while an elite guidance mechanism, selecting superior-performing solutions, ensures consistent convergence toward better solutions. Extensive experiments conducted on the CEC2017 and CEC2022 benchmark functions and practical engineering optimization problems demonstrate that ALBKA achieved the best performance in 85.36% of the 41 tested benchmark functions, representing an absolute improvement of nearly 49 percentage points over the original BKA (36.57%). Additionally, the standard deviation was significantly reduced, indicating improved stability of optimization results. ALBKA exhibits higher solution accuracy and superior practical utility, especially in practical applications such as path planning. These results demonstrate that ALBKA outperforms the original BKA and several advanced swarm intelligence algorithms.
  PubDate: 2025-06-17
Finite-time exponential anti-synchronization of Markovian delayed BAM
neural networks with Dirichlet boundary conditions in the space-time
discretized frame
- Free pre-print version: Loading...
  Rate this result: What is this?
  
  Please help us test our new pre-print finding feature by giving the pre-print link a rating.
  A 5 star rating indicates the linked pre-print has the exact same content as the published article.
  
  Abstract: The finite-time anti-synchronisation in the mean-square sense of time-delayed Dirichlet boundary valued space-time discrete Markovian BAM neural networks is investigated in this paper. In this paper, we derive some results for the criteria of finite-time asymptotic anti-synchronization of BAM neural networks based on the Lyapunov–Krasovskii functional, the discrete inequality of the Wirtinger type and the discrete formula of integration by parts. Incorporating with the Lyapunov–Krasovskii functional involving the double sum of the delay-dependent component, finite-time exponential anti-synchronization is investigated for better showing the synchronized networks than asymptotic anti-synchronization. This study shows that the finite-time mean-squared anti-synchronisation of space-time discrete Markovian BAM neural networks can be better ensured with the smaller diffusion intensities, time delays and larger self-feedback connection weights. In comparison with the studies on global-time anti-synchronisation, the results of this paper, providing a framework to discuss the issue of finite-time anti-synchronisation for space-time discrete models of neural networks, have a wide range of applications. Finally, the validity of the method is verified by means of an illustrative example.
  PubDate: 2025-06-16
Improving multi-hop question answering with prompting explicit and
implicit knowledge aligned human reading comprehension
- Free pre-print version: Loading...
  Rate this result: What is this?
  
  Please help us test our new pre-print finding feature by giving the pre-print link a rating.
  A 5 star rating indicates the linked pre-print has the exact same content as the published article.
  
  Abstract: Language models (LMs) utilize chain-of-thought (CoT) to imitate human reasoning and inference processes, achieving notable success in multi-hop question answering (QA). Despite this, a disparity remains between the reasoning capabilities of LMs and humans when addressing complex challenges. Psychological research highlights the crucial interplay between explicit content in texts and prior human knowledge during reading. However, current studies have inadequately addressed the relationship between input texts and the pre-training-derived knowledge of LMs from the standpoint of human cognition. In this paper, we propose a Prompting Explicit and Implicit knowledge (PEI) framework, which employs CoT prompt-based learning to bridge explicit and implicit knowledge, aligning with human reading comprehension for multi-hop QA. PEI leverages CoT prompts to elicit implicit knowledge from LMs within the input context, while integrating question type information to boost model performance. Moreover, we propose two training paradigms for PEI, and extend our framework on biomedical domain QA to further explore the fusion and relation of explicit and implicit biomedical knowledge via employing biomedical LMs in the Knowledge Prompter to invoke biomedical implicit knowledge and analyze the consistency of the domain knowledge fusion. The experimental results indicate that our proposed PEI performs comparably to the state-of-the-art on HotpotQA, and surpasses baselines on 2WikiMultihopQA and MuSiQue. Additionally, our method achieves a significant improvement compared to baselines on MEDHOP. Ablation studies further validate the efficacy of PEI framework in bridging and integrating explicit and implicit knowledge.
  PubDate: 2025-06-16

HOME > Browse the 73 Subjects covered by JournalTOCs
Subject	Total Journals