This study presents a modified recurrent neural network (RNN) model designed as a parallel computing structure for serial information processing. The result is a novel parallel recurrent neural network (P-RNN), proposed for application to time-varying signal classification. The network uses gated recurrent units (GRUs) for basic information processing and consists of a multi-channel time series signal input layer, parallel processing structure units, a signal feature fusion layer, and a softmax classifier. The P-RNN expands the existing RNN serial processing mode for multi-channel time-varying signals into parallel mode and realizes the embedding of multi-channel signal structure features. In these parallel processing units, the input signal for each channel corresponds to a GRU recurrent network. Feature extraction and attribute association of single-channel signals were performed to achieve parallel processing of all-channel signals. In the feature fusion layer, feature vectors from each channel signal were integrated to generate a comprehensive feature matrix. On this basis, the softmax function was used as a classifier for multi-channel signals. With this mechanism, the P-RNN model achieved independent feature extraction of single-channel signals, characteristic fusion of each channel signal, and signal classification based on an integrated feature matrix. This approach maintained characteristic combination relationships that improved serial modes for existing RNN multi-channel signal processing, reduced the loss of structural feature information, and improved the representation ability of combined feature in local time region and the efficiency of the algorithm. In this paper, the properties of the proposed P-RNN are analyzed and a comprehensive learning algorithm is developed. Seven disease classification types commonly diagnosed using 12-lead ECG signals were used to validate the technique experimentally. Results showed the computational efficiency improved by a factor of 11.519, compared with existing RNN serial processing times, producing a correct recognition rate of 95.976%. In particular, the resolution of signal samples with similar distribution characteristics improved significantly, which demonstrates the effectiveness of the proposed technique.
A novel probabilistic process neural network (PPNN) model is proposed for the multi-channel time-varying signal classification problems with ambiguity and randomness distribution characteristics. This model was constructed from an input time-varying signal layer, a probabilistic process neuron (PPN) hidden layer, a pattern layer, and a Softmax classifier. The number of nodes in the input layer is the same as the number of time-varying signal input channels, which can realize the overall input of the time-varying signal. The PPN hidden layer performs the spatio-temporal aggregation operations for time-varying input signals and probabilistic outputs. The connection weight functions from the input layer to the hidden layer are represented by the typical samples or cluster center functions in different pattern subsets of the sample set, which the prior knowledge of signal categories is implicitly expressed by the morphological distribution features and combinational relations. A pattern layer selectively summed to the output of the PPN hidden layer using the categorical attributes of the connection weight function vector. And the Softmax classifier implements the probabilistic classification of time-varying signals. PPNN has the advantages of less model parameters, suitable for modeling small samples, and integrating prior knowledge of signal categories, etc. This study develops the specific learning algorithms which synthesize dynamic time warping, C-means clustering and BP algorithm. The 12-lead electrocardiogram (ECG) signals for heart disease diagnosis were used for classification testing. Experimental results from 12-lead ECG signal across ten types of disease classification verify the effectiveness of the model and the proposed algorithm.
The existing image semantic segmentation models have low accuracy in detecting tiny targets or multi-targets at overlapping regions. This work proposes a hybrid vision transformer with unified-perceptual-parsing network (ViT-UperNet) for medical image segmentation. A self-attention mechanism is embedded in a vision transformer to extract multi-level features. The image features are extracted hierarchically from low to high dimensions using 4 groups of Transformer blocks with different numbers. Then, it uses a unified-perceptual-parsing network based on a feature pyramid network (FPN) and a pyramid pooling module (PPM) for the fusion of multi-scale contextual features and semantic segmentation. FPN can naturally use hierarchical features, and generate strong semantic information on all scales. PPM can better use the global prior knowledge to understand complex scenes, and extract features with global context information to improve segmentation results. In the training process, a scalable self-supervised learner named masked autoencoder is used for pre-training, which strengthens the visual representation ability and improves the efficiency of the feature learning. Experiments are conducted on cardiac magnetic resonance image segmentation where the left and right atrium and ventricle are selected for segmentation. The pixels accuracy is 93.85%, the Dice coefficient is 92.61% and Hausdorff distance is 11.16, which are improved compared with the other methods. The results show the superiority of Vit-UperNet in medical images segmentation, especially for the low-recognition and serious-occlusion targets.
7 February 2023 in Multimedia Tools and Applications
Graph is an important model to describe various networks, and its scale becomes larger and larger with the development of communication and information technology. The analysis of large-scale graphs requires distributed graph processing systems, and graph partition is the basis of these systems. The existing graph partitioning algorithms are almost proposed for homogeneous clusters, which don't consider the differences among computing nodes in heterogeneous clusters. This paper proposes GAP, a Genetic Algorithm based graph Partitioning algorithm to solve this problem. GAP aims to reduce the total processing time on a heterogeneous cluster by partitioning graphs according to the computing powers of computing nodes. GAP balanced partition the graph initially, and then utilizes genetic algorithm to transfer vertices to reduce cut edges. GAP can balance the processing time of computing nodes, and reduce the communication time among computing nodes. The experiments performed on a heterogeneous cluster demonstrate the outperformance of GAP than Hash.
As a critical technology, anomaly detection ensures the smooth operation of cloud systems while maintaining the market competitiveness of cloud service providers. However, the resource data in real-world cloud systems is predominantly unannotated, leading to insufficient supervised signals for anomaly detection. Moreover, complicated topological associations existed between cloud servers (e.g., computation, storage, and communication). While acquiring resource information, correlating the system topology is challenging. To this end, we propose the GCAD for cloud system anomaly detection, which integrates data augmentation, GraphGRU, contrastive learning, and reconstruction. First, GCAD constructs positive and negative sample pairs through the masking and Gaussian noise data augmentation. Then, the GraphGRU processes extended temporal graph data, extracting and fusing spatiotemporal features from resource status and system topology. In addition, GCAD introduces linear attention for encoding spatiotemporal representations to capture their global correlation information. The weight parameters of the encoder are optimized using a contrastive learning mechanism. Finally, GCAD utilizes a reconstruction technique to calculate anomaly scores, facilitating the evaluation of the state of the cloud system at each time point. Experimental results indicate that GCAD outperforms state-of-the-art compared methods on two real-world datasets that contain topology information.