Aiming at solving the problem that existing artificial neural networks (ANNs) still have low accuracy in predicting yarn strength, this study combines traditional expert experience and an ANN to propose a hybrid network, named the expert weighted neural network. Many studies have shown that it is reliable to predict yarn strength based on ANN technology. However, most ANN training models face with problems of low accuracy and easy trapping into their local minima. The strength prediction of traditional yarns relies on expert experience. Obvious expert experience can help the model perform preliminary learning and help the algorithm model achieve higher accuracy. Therefore, this study proposes a neural network model that combines expert weights and particle swarm optimization (PSO). The model uses PSO to optimize the weights of experts and investigates its effectiveness in yarn strength prediction.
5 December 2025 in CNS Neuroscience & Therapeutics
Objective: The current application of single‐modality electroencephalography (EEG) or functional near‐infrared spectroscopy (fNIRS) to assess consciousness levels in patients with disorders of consciousness (DoC) has garnered significant attention. However, the diagnostic accuracy of unimodal approaches remains suboptimal. Therefore, this study aims to apply the multimodal fusion technology of EEG and fNIRS to the clinical diagnosis of DoC patients. Methods: Eleven patients with DoC (six with a minimally conscious state [MCS] and five with a vegetative state [VS]) were enrolled. The motor intention‐based brain‐computer interface (MI‐BCI) paradigm was adopted. EEG and fNIRS were recorded simultaneously. The synchronous states of EEG and fNIRS were analyzed, including time‐frequency analysis, event‐related desynchronization (ERD), and changes in oxy‐hemoglobin (HbO)/de‐oxygenated (HbR)/total hemoglobin (HbT) content. A multimodal method combining EEG and fNIRS was used to classify DoC patients. Results: The machine‐learning results of the MI‐BCI model showed that the EEG‐fNIRS multimodal approach was superior to single‐modality techniques in the diagnosis of healthy controls (HC), MCS, and VS. The multimodal model achieved a mean AUC of 0.69 ± 0.10, significantly outperforming both unimodal EEG (0.43 ± 0.19; p < 0.01) and standalone fNIRS (0.63 ± 0.10; p < 0.05). The EEG_ERD index of left‐handed MI‐BCI significantly differentiated the MCS and VS groups. Meanwhile, for the classification tasks of HC, MCS, and VS, the importance ranking of the indicators was as follows: fNIRS_ACC, EEG_ACC, fNIRS_slope, fNIRS_centroid, EEG_ERD, fNIRS_integral, and fNIRS_mean. Conclusion: The integration of multimodal MI‐BCI paradigms demonstrates clinical potential in evaluating consciousness levels, while the synergistic combination of neurophysiological and hemodynamic biomarkers provides a robust framework for enhancing the precision of bedside diagnostic protocols. Trial Registration: Clinical Trial Registry: ChiCTR2400085830
Classification models trained on imbalanced datasets tend to be biased towards the majority category, resulting in reduced accuracy for minority categories. A common approach to address this problem is to generate artificial data for underrepresented categories. The Synthetic Minority Over-sampling Technique (SMOTE) algorithm and its variants are widely used for this purpose. In this paper, we propose a modification to the data generation mechanism called Iteration-based SMOTE (ISMOTE). Unlike SMOTE, the ISMOTE algorithm trains the data for multiple iterations. In each iteration, the model generates new samples in the vicinity of appropriately misclassified data. These new samples are then fed into the classification model, thus improving classification accuracy over the course of multiple iterations. We compare the performance of ISMOTE with SMOTE and other commonly used oversampling algorithms. Our empirical results demonstrate that ISMOTE significantly improves the quality of the generated data compared to other oversampling methods. Additionally, we conduct experiments to verify the effect of parameters on the model and provide suggestions for choosing appropriate values to improve performance.
28 July 2021 in Journal of The Institution of Engineers (India): Series E
With the continuous development of deep learning, due to the complexity of the deep neural network structure and the limitation of training time, some scholars have proposed broad learning, the Broad Learning System (BLS). However, BLS currently only verifies that it has excellent effects on some of the network training data sets, and it does not necessarily have excellent effects on some actual data sets. In response to this, this paper uses the effect of BLS in predicting the unevenness of yarn quality in the yarn data set, and proposes a BLS-based multi-layer neural network (MNN) for the problems, which is called Broad Multilayer Neural Network (BMNN).