Soft Computing最新文献

In response to growing security concerns and the increasing demand for face recognition (FR) technology in various sectors, this research explores the application of deep learning techniques, specifically pre-trained Convolutional Neural Network (CNN) models, in the field of FR. The study harnesses the power of five pre-trained CNN models—DenseNet201, ResNet152V2, MobileNetV2, SeResNeXt, and Xception—for robust feature extraction, followed by SoftMax classification. A novel weighted average ensemble model, meticulously optimized through a grid search technique, is introduced to augment feature extraction and classification efficacy. Emphasizing the significance of robust data pre-processing, encompassing resizing, data augmentation, splitting, and normalization, the research endeavors to fortify the reliability of FR systems. Methodologically, the study systematically investigates hyperparameters across deep learning models, fine-tuning network depth, learning rate, activation functions, and optimization methods. Comprehensive evaluations unfold across diverse datasets to discern the effectiveness of the proposed models. Key contributions of this work encompass the utilization of pre-trained CNN models for feature extraction, extensive evaluation across multiple datasets, the introduction of a weighted average ensemble model, emphasis on robust data pre-processing, systematic hyperparameter tuning, and the utilization of comprehensive evaluation metrics. The results, meticulously analyzed, unveil the superior performance of the proposed method, consistently outshining alternative models across pivotal metrics, including Recall, Precision, F1 Score, Matthews Correlation Coefficient (MCC), and Accuracy. Notably, the proposed method attains an exceptional accuracy of 99.48% on the labeled faces in the wild (LFW) dataset, surpassing erstwhile state-of-the-art benchmarks. This research represents a significant stride in FR technology, furnishing a dependable and accurate solution fortified by empirical substantiation. The proposed method showcases the potential of pre-trained CNN models, ensemble learning, robust data pre-processing, and hyperparameter tuning in augmenting the accuracy and reliability of FR systems, with far-reaching implications for real-world applications.

Decomposition structure learning algorithms are widely adopted to recover Bayesian network structures. In the recursive process of separation phase, the network partition is obtained through recursively two steps: constructing the undirected independence graph (UIG) and decomposing with the help of partition methods. UIG as the basis for decomposition directly affects the result of the network partition and then impacts the accuracy of output structure. Existing construction algorithms adopt a fixed type of UIG in the recursive process and researches divide into two directions: constructing moral graph and moral graph with extra edges. The former suffer from the problem that computational complexity of recovering all conditional independences (CIs) is too high to divide network well due to relatively complex networks at the beginning of the recursive process, while the latter suffer from the problem that the network partition is hard to find by insufficient expression degree of CIs due to relatively simple networks at the end of the recursive process. The reason is that the fixed type of UIG can not cope with variation of network size. Therefore, this paper proposes a stage-driven construction algorithm considering variation of network size in the recursive process. Different from other construction algorithms, the proposed algorithm designs the network scale factor to achieve the stage division of the recursive process, and selects different algorithms at different stages to build appropriate UIGs through demand analysis. Experiments on different benchmark networks verify that the proposed algorithm can obtain better performances compared with other representative algorithms.

Flue-cured tobacco (FCT) can be classified into upper (B), middle (C), and lower (X) parts based on characteristics such as the FCT's main veins, leaf shape, color, and thickness. Accurately measuring the geometric parameters of the main veins is crucial for identifying the different parts. However, this task has proven to be challenging. Therefore, segmenting the main veins is a prerequisite to reducing calculation errors and improving the precision of part identification. To obtain enough semantic information and improve segmentation accuracy, we propose a fine segmentation model (MSHF-Net) of FCT's main veins based on multi-level-scale features of hybrid fusion. Firstly, MobileNetV2 with a dilated convolution layer (DMobileNetV2) is selected as the backbone network for feature extraction, which optimizes training and inference speed to minimize computing costs. Subsequently, Hybrid Fusion Atrous Spatial Pyramid Pooling (HFASPP) is designed to be the strengthened backbone module for capturing more high-level semantic information, effectively preventing intermittent segmentation of some main veins. Additionally, considering the low proportion of main vein targets in the original image, the double shallow feature branches (DSFBS) are included to obtain more low-level semantic information. Finally, a channel attention mechanism (ECANet) is added to enhance useful information and eliminate redundant information after the hybrid fusion of high-low-level semantic information, preventing mis-segmentation of regions. Experimental validation demonstrates the efficiency of the MSHF-Net, with parameters of only 7.92 M, thus ensuring minimal computational requirements. The model achieves an impressive mean intersection over union (MIoU) of 85.57% and mean pixel accuracy (mPA) of 93.10% on a diverse test set of FCT parts. When applied to segment main veins in a 2296 × 1548 × 3 tobacco image, the model takes just over 0.1 s. It is noteworthy that none of the 291 randomly segmented tobacco leaf main veins show mis-segmentation, highlighting the model's robustness and practical applicability in various scenarios. These results emphasize the superior segmentation performance of the proposed model, establishing a crucial foundation for accurately discriminating FCT parts.

Although the concept of circular economy is a frequently encountered effective tool of sustainable development, its social dimension, the social circular economy, is a topic that has only begun to be discussed. Assessing the social circular economy performance of European Union countries from the sustainable development perspective is critical for monitoring their progress. In this study, a model that evaluates the social circular economy performances of countries for sustainable development is proposed. The importance weights of the evaluation criteria are determined using the interval-valued intuitionistic fuzzy VIKOR method’s linguistic scale. Then , the countries are ranked using the grey relational analysis method. In this study, the social circular economy performance of EU member states for 2021 are investigated with the proposed model. According to the results obtained, social circular economy performances are directly proportional to the success of countries in achieving sustainable development goals. The success rating achieved in this study may vary with the improvement activities of the relevant countries. The proposed model updates the ranking, considering each improvement. The findings of this study can help scholars and policymakers better understand the social circularity capabilities of the European Union member countries in the context of sustainable development. In the limited literature, there is no other study in which the social circular economy performance of EU member countries is measured with the relevant evaluation criteria. A model that can evaluate the social circular economy performances of the European Union member countries is proposed to fill the profound gap in this field. We contributed to the literature with a new model that differs with evaluation criteria, real and current data sets.

Yang introduced binary MICA (Monotonic Identity Commutative Aggregation) operations called micanorms and micanorms with three weak forms of associativity. This paper investigates general and specific structure of those micanorms. For this, we first recall (weakly associative) micanorms introduced by Yang. Next, we introduce a construction to characterize those micanorms in general. Finally, we consider similar constructions to specify them together with some examples to illustrate those constructions.

In recent years, more and more cryptocurrency markets have adopted Internet of things (IoT) technology, with adoption rates steadily increasing since 2020. Industries that have seen significant growth in adoption include manufacturing, transportation, and public spaces. IoT connects all objects to the Internet through information sensing devices, enabling information exchange and intelligent identification and management of things. This has led to a wide range of applications. While blockchain technology addresses the centralization issue of traditional IoT systems, privacy leakage remains a key challenge for implementing smart supply chain management using IoT. We present privacy-preserving techniques to enable secure IoT trading. First, we propose a blockchain architecture based on post-quantum cryptography to securely store data and information. Second, we propose a post-quantum mixed currency mechanism for enhanced privacy protection. Third, we propose a blockchain-based IoT architecture designed for secure trading. By integrating blockchain, mixed currency approaches, and post-quantum techniques, we develop a blockchain-based IoT trading system. We implement this system within cryptocurrency markets. Our implementation and comparison with related solutions demonstrate that the system provides secure services for IoT trading.

This article proposes a novel idea for creating a sentiment-based stock market index forecasting model by amalgamating price and sentiment data hidden in the price pattern itself. The state-of-the-art methodologies used in forecasting stock markets involve gathering sentiment data from external sources like tweets, but the proposed model is unique in the sense it extracts the sentiment information from the price itself, making it more reliable and easier to test and implement. In the proposed system the simple daily time series is converted to an information enriched fuzzy linguistic time series with a unique approach of incorporating information about the sentiment behind the Open High Low Close (OHLC) price formation that took place at every instance of the time series. A unique approach is followed while modeling the information retrieval (IR) system which converts a simple IR system it into a forecasting system. A number of experiments were conducted using the proposed model on Nifty-50 index values (5 years) and it was found that the Root Mean Squared Error (RMSE) value came around 1.03 and RMSE% came around 1.72% which is quite small compared to number of observations and hence this gives a strong indication that the proposed system has the capability to perform good quality of forecasts. The model is simple and easy to implement with very nominal memory requirements, compared to other type of models.

Stream mining, especially with concept drift, presents significant challenges across various domains. As data streams evolve over time, initial models become less effective. We present a novel approach using fuzzy ARTMAP’s adaptability and neural networks’ robustness to address concept drift. Our method dynamically updates models based on changing data distributions, enabling real-time adap- tation. By integrating fuzzy ARTMAP with backpropagation, it facilitates agile learning and accurate predictions in evolving scenarios. Through rigorous exper- iments, we demonstrate the effectiveness of our method in managing concept drift and achieving substantial performance improvements. The achieved accu- racy of 85.07% and F1 score of 72.47 demonstrate the effectiveness of the approach in real-time classification tasks. This research extends beyond just performance metrics. By leveraging the interpretability of fuzzy ARTMAP, we gain valuable insights into the mechanisms that enable our model to adapt to concept drift. This deeper understanding paves the way for further advancements in this area.

A real-time wind velocity vector and parameters estimation and wind model identification approach using a machine learning technique is addressed in this paper. The proposed method uses only the state measurements of an aircraft and does not require control commands, air data systems, or satellite-based data. Small unmanned aerial vehicles (UAVs) can benefit from this method, since it relies solely on measurement results from the common sensors as an attitude and heading reference system. The independence of external sources of information made estimations resistant to intentional errors. This algorithm uses long short-term memory neural networks (LSTM NNs) in a two-step deep learning process involving classification and regression. A classification NN was trained with four different labeled wind models, while individual regression NNs were trained to estimate the velocity vector and parameters of each wind model. The linear acceleration, angular velocity, and Euler angle measurements were used as the inputs of trained networks. The algorithm suggests in its first step identifying the exact wind model, and in its second step estimating the wind velocity vector and parameters using a properly assigned estimation from a trained network. A nonlinear six-degree-of-freedom simulation of straightforward and level turn maneuvers of a fixed-wing UAV in the presence of different wind models served as the dataset in the learning process. Monte Carlo simulations proved the accuracy and rapidity of the proposed algorithm in identifying the wind model and estimating three-dimensional wind velocity vector and parameters.

The effective prediction of heart disorders is crucial for timely intervention and treatment before a cardiac event occurs. While various machine learning models have been developed for this purpose, many struggle to handle high-dimensional data effectively, limiting their performance. In this work, efforts have been made to enhance the performance and computational efficiency of deep learning classifiers using hyperparameters. The study utilized heart sound data from normal and diseased patients obtained from standard online repositories. The hyperparameter tuned modified CNN-based Inception Network model achieved an accuracy of 99.65% ± 0.23% on the test dataset, along with a sensitivity of 98.8% ± 0.12% and specificity of 98.2% ± 0.15%. Thus the hyperparameter-tuned CNN-based Inception Network model outperformed its counterparts, making it the most effective model for diagnosing heart disorders.