Jie Cheng

The integration of deep learning methods, especially Convolutional Neural Networks (CNN), and Synthetic Aperture Radar Automatic Target Recognition (SAR ATR) has been widely deployed in the field of radar signal processing. Nevertheless, these methods are frequently regarded as black-box models due to the limited visual interpretation of their internal feature representation and parameter organization. In this paper, we propose an innovative approach named C-RISE, which builds upon the RISE algorithm to provide a post-hoc interpretation technique for black-box models used in SAR Images Target Recognition. C-RISE generates saliency maps that effectively visualize the significance of each pixel. Our algorithm outperforms RISE by clustering masks that capture similar fusion features into distinct groups, enabling more appropriate weight distribution and increased focus on the target area. Furthermore, we employ Gaussian blur to process the masked area, preserving the original image structure with optimal consistency and integrity. C-RISE has been extensively evaluated through experiments, and the results demonstrate superior performance over other interpretation methods based on perturbation when applied to neural networks for SAR image target recognition. Furthermore, our approach is highly robust and transferable compared to other interpretable algorithms, including white-box methods.

The integration of deep learning methods, especially Convolutional Neural Networks (CNN), and Synthetic Aperture Radar Automatic Target Recognition (SAR ATR) has been widely deployed in the field of radar signal processing. Nevertheless, these methods are frequently regarded as black-box models due to the limited visual interpretation of their internal feature representation and parameter organization. In this paper, we propose an innovative approach named C-RISE, which builds upon the RISE algorithm to provide a post-hoc interpretation technique for black-box models used in SAR Images Target Recognition. C-RISE generates saliency maps that effectively visualize the significance of each pixel. Our algorithm outperforms RISE by clustering masks that capture similar fusion features into distinct groups, enabling more appropriate weight distribution and increased focus on the target area. Furthermore, we employ Gaussian blur to process the masked area, preserving the original image structure with optimal consistency and integrity. C-RISE has been extensively evaluated through experiments, and the results demonstrate superior performance over other interpretation methods based on perturbation when applied to neural networks for SAR image target recognition. Furthermore, our approach is highly robust and transferable compared to other interpretable algorithms, including white-box methods.

In this paper we study simulation-based methods for estimating gradients in stochastic networks. We derive a new method of calculating weak derivative estimator using importance sampling transform, and our method has less computational cost than the classical method. In the context of M/M/1 queueing network and stochastic activity network, we analytically show that our new method won't result in a great increase of sample variance of the estimators. Our numerical experiments show that under same simulation time, the new method can yield a narrower confidence interval of the true gradient than the classical one, suggesting that the new method is more competitive.

Making an accurate prediction of an unknown system only from a short-term time series is difficult due to the lack of sufficient information, especially in a multi-step-ahead manner. However, a high-dimensional short-term time series contains rich dynamical information, and also becomes increasingly available in many fields. In this work, by exploiting spatiotemporal information (STI) transformation scheme that transforms such high-dimensional/spatial information to temporal information, we developed a new method called MT-GPRMachine to achieve accurate prediction from a short-term time series. Specifically, we first construct a specific multi-task GPR which is multiple linked STI mappings to transform high dimensional/spatial information into temporal/dynamical information of any given target variable, and then makes multi step-ahead prediction of the target variable by solving those STI mappings. The multi-step-ahead prediction results on various synthetic and real-world datasets clearly validated that MT-GPRMachine outperformed other existing approaches.

In order to solve the problem of low security of data in network transmission and inaccurate prediction of future security situation, an improved neural network learning algorithm is proposed in this paper. The algorithm makes up for the shortcomings of the standard neural network learning algorithm, eliminates the redundant data by vector support, and realizes the effective clustering of information data. In addition, the improved neural network learning algorithm uses the order of data to optimize the "end" data in the standard neural network learning algorithm, so as to improve the accuracy and computational efficiency of network security situation prediction.MATLAB simulation results show that the data processing capacity of support vector combined BP neural network is consistent with the actual security situation data requirements, the consistency can reach 98%. the consistency of the security situation results can reach 99%, the composite prediction time of the whole security situation is less than 25s, the line segment slope change can reach 2.3% ,and the slope change range can reach 1.2%,, which is better than BP neural network algorithm.

Due to the combination of the forward speed and the prevailing wind for surface ship traveling in the ocean, the airflow passing over the ship’s superstructure causes the formation of a disturbed flow region and the large speed gradients of the mean wind over the flight deck, known as the ship airwake. This airwake would cause significant influence on the performance of the helicopter rotor during its taking off or landing, increase the operation workload of the pilot and even cause safe-landing issues, especially when the wind sweeps over the deck. This paper presents a numerical simulation of flow across the ship superstructure using DES and LES turbulent model. The ship model used for simulation is the standard SF2 surface ship model with experimental measurement data which could be used for the CFD code validation. The simulation results are compared with the experimental measurement data, and the comparison with experimental results shows good match for both DES and LES turbulent models. Simulation results show that a series of vortex had been generated after the flow separation with asymmetric characteristics. From upstream to downstream, the vortex intensity decreases, but suddenly increases after encountering the chimney. The comparison between DES and LES turbulent models shows the similar flow field and vortex structure around the ship superstructure with same grid sets. Both DES and LES are superior to RANS in solving ship airwake. The comparisons of DES and LES turbulent models show that DES can reflect the separated flow with limited computational resource and LES simulation could get higher resolution of the fluid flow structure with enough computational resources.