Epilepsy is a prevalent and unpredictable neurological disorder that requires EEG analysis for diagnosis. Given the non-stationary nature of EEG recordings, advanced signalprocessing techniques are essential for clas...
详细信息
ISBN:
(纸本)9798350351491;9798350351484
Epilepsy is a prevalent and unpredictable neurological disorder that requires EEG analysis for diagnosis. Given the non-stationary nature of EEG recordings, advanced signalprocessing techniques are essential for classifying epilepsy EEG signals accurately into normal, pre-ictal, and ictal signals. Recent research has suggested using temporal frequency representations like Spectrogram, Scalogram, Smoothed Pseudo Wigner-Ville distribution (SPWVD), and Hilbert-Huang Transform (HHT). In this study, we proposed a computerized classification method for epileptic EEG signals. We conduct a comparative analysis of nonstationary signalprocessing techniques alongside a deep convolutional neural network (CNN) classifier. Our research utilizes the EEG database from the University of Bonn. Our findings indicate variations in classification results among the employed feature extraction methods. Spectrogram techniques, in particular, show promising results with an average accuracy of 98.66%.
This paper emphasizes on detection and classification of Fluoro-Deoxy-Glucose (FDG) radioactivity uptakes in fused Positron Emission Tomography / Computerized Tomography (PET/CT) images automatically. The deep learnin...
详细信息
This paper emphasizes on detection and classification of Fluoro-Deoxy-Glucose (FDG) radioactivity uptakes in fused Positron Emission Tomography / Computerized Tomography (PET/CT) images automatically. The deep learning technique using Convolutional neural Network (CNN) is proposed to reduce the complexity in observation, to solve the problem of low accurateness and the time-consuming process of traditional classification methods. The CNN layers are designed and proposed for the FDG uptakes classification problem in fused PET/CT images. The proposed modified CNN model is trained using different optimizers such as stochastic Gradient Descent Momentum (SGDM), Adaptive Moment Estimation (ADAM), and Root Mean Square propagation (RMSprop). The deep features extracted from the proposed CNN are classified using different classifiers such as K-Nearest Neighbor (KNN), Decision Trees (DT), Ensemble, Naive Bayes (NB), and multi-class Support Vector Machine (SVM) the results of which are compared. The multi-class SVM classifier trained using SGDM optimizer attains the maximum test accuracy of 98.18% and was found to be superior to pre-trained deep models such as AlexNet, ResNet, and GoogleNet.
Electrical Impedance Tomography (EIT) is a highly ill-posed inverse problem, with the challenge of reconstructing internal conductivities using only boundary voltage measurements. Although Physics-Informed neural Netw...
详细信息
Electrical Impedance Tomography (EIT) is a highly ill-posed inverse problem, with the challenge of reconstructing internal conductivities using only boundary voltage measurements. Although Physics-Informed neural Networks (PINNs) have shown potential in solving inverse problems, existing approaches are limited in their applicability to EIT, as they often rely on impractical prior knowledge and assumptions that cannot be satisfied in real-world scenarios. To address these limitations, we propose a two-stage hybrid learning framework that combines Convolutional neural Networks (CNNs) and PINNs. This framework integrates data-driven and model-driven paradigms, blending supervised and unsupervised learning to reconstruct conductivity distributions while ensuring adherence to the underlying physical laws, thereby overcoming the constraints of existing methods.
Mixture of experts with a sparse expert selection rule has been gaining much attention recently because of its scalability without compromising inference time. However, unlike standard neural networks, sparse mixture-...
详细信息
Mixture of experts with a sparse expert selection rule has been gaining much attention recently because of its scalability without compromising inference time. However, unlike standard neural networks, sparse mixture-of-experts models inherently exhibit discontinuities in the output space, which may impede the acquisition of appropriate invariance to the input perturbations, leading to a deterioration of model performance for tasks such as classification. To address this issue, we propose Pairwise Router Consistency (PRC) that effectively penalizes the discontinuities occurring under natural deformations of input images. With the supervised loss, the use of PRC loss empirically improves classification accuracy on imageNet-1 K, CIFAR-10, and CIFAR-100 datasets, compared to a baseline method. Notably, our method with 1-expert selection slightly outperforms the baseline method using 2-expert selection. We also confirmed that models trained with our method experience discontinuous changes less frequently under input perturbations.
This paper discusses the sampled-data input-to-state stabilization for delayed semi-Markovian jump neural networks subject to external disturbance. First, a hybrid closed-loop system is formulated, which contains cont...
详细信息
This paper discusses the sampled-data input-to-state stabilization for delayed semi-Markovian jump neural networks subject to external disturbance. First, a hybrid closed-loop system is formulated, which contains continuous-time state signal, disturbance input signal, discrete-time control signal, and jumping parameters of the semi-Markovian process. Then, two time-dependent and mode-dependent Lyapunov functionals are constructed corresponding to different assumptions about the activation functions. Subsequently, two sufficient conditions concerning the sampled-data controller design are derived to ensure the mean-square input-to-state stability for the hybrid closed-loop system by utilizing the proposed Lyapunov functionals, a few inequalities, as well as some stochastic analysis techniques. It is worth remarking that the present conditions are capable of ensuring mean-square exponential stability of the closed-loop system in the absence of external disturbances. Lastly, a numerical example is employed to verify the validity of the proposed input-to-state stabilization methods.
The neurocomputing communities have focused much interest on quaternionic-valued neural networks (QVNNs) due to the natural extension in quaternionic signals, learning of inter and spatial relationships between the fe...
详细信息
The neurocomputing communities have focused much interest on quaternionic-valued neural networks (QVNNs) due to the natural extension in quaternionic signals, learning of inter and spatial relationships between the features, and remarkable improvement against real-valued neural networks (RVNNs) and complex-valued neural networks (CVNNs). The excellent learning capability of QVNN inspired the researchers working on various applications in imageprocessing, signalprocessing, computer vision, and robotic control system. Apart from its applications, many researchers have proposed new structures of quaternionic neurons and extended the architecture of QVNN for specific applications containing high-dimensional information. These networks have revealed their performance with a lesser number of parameters over conventional RVNNs. This paper focuses on past and recent studies of simple and deep QVNNs architectures and their applications. This paper provides the future directions to prospective researchers to establish new architectures and to extend the existing architecture of high-dimensional neural networks with the help of quaternion, octonion, or sedenion for appropriate applications.
We extensively survey applications of Clifford Geometric algebra in recent years (mainly 2019-2022). This includes engineering;electric engineering;optical fibers;geographic information systems;geometry;molecular geom...
详细信息
We extensively survey applications of Clifford Geometric algebra in recent years (mainly 2019-2022). This includes engineering;electric engineering;optical fibers;geographic information systems;geometry;molecular geometry;protein structure;neural networks;artificial intelligence;encryption;physics;signal, image, and video processing;and software.
Extracting robust feature representation is one of the key challenges for person re-identification (ReID) task. Although convolution neural network (CNN)-based methods have achieved great success, they still cannot ha...
详细信息
Extracting robust feature representation is one of the key challenges for person re-identification (ReID) task. Although convolution neural network (CNN)-based methods have achieved great success, they still cannot handle the part occlusion and misalignment caused by limited receptive field. Recently, pure transformer models have shown its power in the person ReID task. However, current transformer models adopt patches of equal-scale as input, and cannot solve the problem of cross-scale interaction properly. To overcome this problem, an adaptive cross-scale transformer from a perspective of the graph signal, named ACSFormer, is proposed. Specifically, the self-attention module is first treated as an undirected fully connected graph. And then, "node variation" is introduced as an indicator to adaptively merge neighbourhood tokens. To the best of the authors' knowledge, their ACSFormer is the first work to attempt to combine pure transformers and graph signalprocessing in the field of person ReID. Extensive evaluations are conducted on three person ReID datasets to validate the performance of ACSFormer. Experiments demonstrate that this ACSFormer performs on par with state-of-the-art CNN-based methods and consistently improves transformer-based baseline, for example, surpassing ViT-baseline by 2.5%, 2.7% and 4.8% mAP on Market1501, DukeMTMC-reID and MSMT17, respectively.
In recent years, significant progress has been made in image compression sensing (ICS) through deep learning techniques. Deep Unfolding Networks (DUN) transforms the iterative reconfiguration process into an end-to-en...
详细信息
In recent years, significant progress has been made in image compression sensing (ICS) through deep learning techniques. Deep Unfolding Networks (DUN) transforms the iterative reconfiguration process into an end-to-end deep neural network, improving interpretability and performance. However, traditional algorithms are limited to processing information in pixel space, missing the potential advantages of feature space. Additionally, most DUN are constrained by fixed input-output mirror structures that restrict information flow and lack adaptability due to their use of a fixed threshold for soft shrinkage operations. To address these limitations, we propose a novel feature space-based compression-aware adaptive threshold network (FNAT-Net). The supplementary information (FI) is utilized to enable FNAT-Net to perform fusion processing across both the pixel and feature domains, mapping a two-step approximate gradient descent algorithm from pixel to feature space. Furthermore, this paper introduces an effective enhanced Multi-Layer Perceptron (MLP) adaptive soft-thresholding strategy. This strategy enables FNAT-Net to address L1-regularized neighbourhood mappings with content-aware thresholds. FNAT-Net outperforms state-of-the-art methods, demonstrating superior performance across a wide range of scene changes and noise conditions.
In the field of digital imageprocessing, image segmentation technology has always been a key technology studied and discussed by many scholars. There may be certain contradictions in the implementation of image segme...
详细信息
In the field of digital imageprocessing, image segmentation technology has always been a key technology studied and discussed by many scholars. There may be certain contradictions in the implementation of image segmentation technology, namely the contradiction between segmentation accuracy and difficulty, as well as the contradiction between excessive segmentation and deficient segmentation. These contradictions make it difficult for fixed scale segmentation techniques to achieve perfect image segmentation. This article utilizes artificial neural networks and related genetic algorithms, based on optimization theory as a prototype, to establish gamut set intervals for digital imaging. It deeply explores and studies the application of optimization theory between device feature models and gamut matching models, and obtains several forward and backward feature models, and evaluates and compares them. By studying the specific methods of multi-scale theoretical image segmentation technology, it was found that using wavelet data transformation methods can perform multi-level preprocessing on images, and combining multi-scale autoregressive models to achieve image recognition and segmentation. After constructing multi-scale data structures, multi-scale segmentation is implemented, and Markov stochastic models are used for strategic research in image segmentation. The final experiment shows that the algorithm proposed in this article has excellent speed and accuracy.
暂无评论