Search

• The departure delay is predicted based on node sequence data of ground support services for transit flights. • The cases of Shanghai Pudong and Kunming Changshui airports are applied for model illustration. • A Bayesian network is constructed for interrelationship analysis between segments. • A GCN-GraphSAGE model is proposed for departure delay prediction. • The proposed GCN-GraphSAGE method is reported to outperform the other benchmark methods. Delays frequently occur during ground support services for transit flights. Due to the limitation of related data acquisition, the delay occurred during this process has seldom been explored. Toward the purpose of departure delay mitigation, the present research provides the node sequence data to identify the critical ground support service nodes that associate with departure delay. Machine learning techniques are used for accurate departure delay prediction of transit flights. First, the topological structure of the neural network is constructed according to the ground service support procedure of transit flights. The flight data from Shanghai Pudong and Kunming Changshui International airports are collected for delay description and model evaluation. Then, similarity of delay scenarios are illustrated and the typical delay patterns are explored based on clustering methods. It is found that there are some node sections that delays routinely occur. Third, a Bayesian network is constructed for delay probability analysis. The conditional probability table describing the impact of delay occurred during each section on the departure delay for transit flights is provided. Finally, the performance of two classes of approaches to predict departure delays are constructed and compared, including three traditional machining learning methods denoted as random forest, decision tree, XGBoost models, and four graph convolutional neural network models namely Graph convolutional network (GCN), Graph attention network (GAT), Graph sample and aggregate (GraphSAGE) and integrated Graph convolutional network and Graph sample and aggregate (GCN-GraphSAGE) models. In general, the proposed GCN-GraphSAGE method achieves the most stable and satisfactory performance for departure delay prediction. The research results can support the dynamic and accurate prediction of departure delays and the refined management of transit flights for delay alleviation. [ABSTRACT FROM AUTHOR]

• A graph-to-sequence DL framework is proposed to predict network-wide flight delay. • The proposed AG2S may capture complex spatiotemporal dynamics in flight delay variables. • Four network-wide flight delay patterns are clustered by K-means analysis. • Graph network analysis is conducted to uncover the black box of the proposed AG2S-Net. • The proposed AG2S-Net outperforms some state-of-the-art models. The primary objective of this study is to predict network-wide multi-step-ahead flight delay. A novel graph-to-sequence learning architecture with attention mechanism (AG2S-Net) is developed to predict the multi-step-ahead hourly departure and arrival delay of the entire network. The proposed AG2S-Net consists of a graph convolution neural network, a bi-LSTM neural network, and a sequence-to-sequence framework with embedded attention mechanism. Five-year flight data of 75 airports and 242 links are collected from the National Airspace System to illustrate the procedure. First, K-means clustering algorithm is applied to classify the hourly flight delay states of the entire network into four typical delay patterns, which are further used as explanatory variables in prediction model. Then, the network-wide multi-step-ahead flight delay prediction model is built with the proposed AG2S-Net. The result indicates that the model achieves better performance on large-scale airports than small-scale airports. The delay pattern variables and weather variables can greatly improve the prediction performance. In addition, compared with several benchmark methods, the AG2S-Net performs the best for all three airport types in terms of the lowest RMSE and MAE values. Finally, the graph network analysis further reveals that the proposed AG2S-Net can capture the hidden correlations between airports without links and collect information from more airports in the same community to enhance the prediction performance. The results of this study could provide insightful suggestions for aviation authorities and airport regulators to develop effective air traffic control strategies for alleviating flight delays and improving operation efficiency of the entire network. [ABSTRACT FROM AUTHOR]

The gap between Neural Architecture Search (NAS) and hardware embedded accelerators degrades the deployment efficiency, due to the absent of rethinking the applicability of the searched network layer characteristics and hardware mapping. Therefore, this work proposes a novel hardware-aware NAS framework in consideration of a deduced efficiency metric. Beside, a layer adaptive scheduler and a coarse-grained reconfigurable computing architecture are developed to deploy the searched networks by selecting the most appropriate dataflow pattern layer-by-layer. Evaluation results show that the proposed NAS framework can search networks with high accuracy and low inference latency, and the proposed architecture provides power-efficiency improvement. • Bridge the applicability gap between NAS networks and hardware embedded accelerators. • A novel hardware-aware NAS framework that incorporates a deduced efficiency metric. • A layer adaptive scheduler that generates the optimal dataflow and control sequence. • A flexible coarse-grained reconfigurable architecture to deploy searched networks. [ABSTRACT FROM AUTHOR]

• A multi-view graph to sequence learning method is built to predict metro ridership. • The AW-MV-G2S model can uncover spatiotemporal dependencies between stations. • Three different types of adjacency matrixes are designed in the developed model. • The developed AW-MV-G2S model shows great transferability in other metro system. • The developed AW-MV-G2S model outperforms some state-of-the-art benchmark models. The primary objective of this study is to forecast network-wide multi-step metro ridership with a novel attention-weighted multi-view graph to sequence learning approach (AW-MV-G2S). The developed AW-MV-G2S model employs multiple graph convolutional neural networks to capture spatial heterogeneous correlations between stations from geographic distance view, functional similarity view and demand pattern view, respectively. A bidirectional LSTM neural network and the attention mechanism is utilized to encode the long-range temporal dependencies in multiple time steps. A three-month trip record of 64 stations on four metro lines is collected from Nanjing Metro System to validate the model. The results indicate that the developed AW-MV-G2S model can fully encode the spatiotemporal characteristics in network-wide metro ridership data, and achieve better prediction accuracy and more robust performance than other compared models when making predictions across multiple look-ahead time steps for all three metro station types. Moreover, the model transferability result also reveals that the developed multi-view graph-to-sequence learning framework can be well transferred to other metro systems with various network structures. The results of this study can help the metro system authorities to dynamically modify the operation plans according to the fluctuation of passenger flow, such as adjusting the headway and train dispatching schedule to ensure the service quality of the entire metro system. [ABSTRACT FROM AUTHOR]

As the deep neural networks (DNNs) with satisfied expression ability usually require large scale to gain adequate performance, and deploying large-scale DNNs on resource-limited environments is still a challenge, neural network compression becomes a hot topic nowadays. Among the multiple compression methods, tensor decomposition reveals many specific advantages, such as regular data structure comes from linear algebra, convenient approach of training from scratch, ideal compression ratio, etc. Nevertheless, for some compact neural modules such as two-dimensional/three-dimensional (2D/3D) convolutional kernels, traditional tensor decomposition in the way of approximation has encountered intractable obstacles. Fortunately, some works utilize the tensor mapping approach, which just regards the data structure of tensor decomposition as neural layers, to reconstruct the convolutional kernel into a new lightweight module with several thinner convolutional kernels. The only two flies in the ointment are there lack necessary theories of tensor mapping, and there is still no tensor mapping way to compress high-order three-dimensional convolutional neural networks (3DCNNs). In this paper, we first deeply analyze the tensor mapping theory including convergence and precision, which separately establishes the rationality of tensor mapping and its superiority over the traditional tensor approximation, according to the Lottery Ticket Hypothesis. Then we propose an efficient method termed as 3D-KCPNet, to compress 3DCNNs based on the novel Kronecker canonical polyadic (KCP) tensor decomposition. The proposed method can not only compress the 3D convolutional kernels, but also convert a 3D convolution to efficient 1 × 1 × 1 and 2D depthwise convolutions. The experiments on the video recognition datasets VIVA Challenge, UCF11, UCF50, and UCF101 show that the accuracy of 3D-KCPNet can surpass its original baseline model and the corresponding tensor approximation model. [ABSTRACT FROM AUTHOR]

We introduce the convolutional neural network (CNN) to extract the effective information of some complex signals in the fiber sensing, and verify the feasibility in the specific process of fiber Bragg grating (FBG) demodulation. In the experiments, in order to eliminate the cross interference in the energized coil, classical double grating configuration is adopted to distinguish magnetic field and temperature. At the same time, the introduction of thermal induced chirp (TIC) phenomenon produces more wavelength components to achieve more precise demodulation. The signal received by oscilloscope presents complex waveform, and the traditional analytical solution cannot demodulate it precisely, while the deep learning capabilities of neural networks can be leveraged to solve this problem. The one-dimensional (1D) time-series signal is transformed into a two-dimensional (2D) image by using the Gramian angle field (GAF) method. On this basis, the data are augmented according to the characteristics of signal noise. Five classical neural network structures are used to process the data. The root mean square errors (RMSEs) of the demodulation results can achieve 0.3 °C and 17.1 Gs respectively, which are relatively high precision in terms of intensity demodulation. This method can be extended to demodulate 1D frequency domain signals and wavelength domain signals, rendering a demodulation scheme to process some complex signals in optical fiber sensing effectively. • A novel FBG intensity demodulation method which can better extract features of complicated signals. • The thermal induced chirp of the directly modulated laser is used to conveniently generate more wavelength components. • GAF transformation converts complicated 1D signals into 2D image while preserving temporal features. • Demonstrate that it is feasible to treat classification task outputs as weights to calculate regression variables. [ABSTRACT FROM AUTHOR]