The use of adversarial training enables rapid training and the isolation of anomalies. This method uses the automatic encoder architecture to meet the conditions of unsupervised learning. proposed a fast and stable unsupervised anomaly detection method, USAD, for multivariate time series. The VAE structure aims to capture the structural rules of the time subsequence on the local window, while the LSTM structure models the changing trend of the long-term time series. proposed a mixed anomaly detection method, which combines the representation learning ability of the VAE with the time modeling ability of the LSTM. The model jointly trains the encoder, generator, and discriminator, which can improve the fidelity of signal reconstruction, make the distinction between normal and anomalies more significant, and improve anomaly detection Accuracy. This method not only reached a very high level in the public dataset but also gave a theoretical inference to transform the model into a Bayesian network, which enhanced the interpretability of the model. used the confrontational training method, Buzz, to detect anomalies in complex time series. constructed the donut algorithm based on the VAE, which trains the normal and abnormal data simultaneously, making the feature extraction more complete and providing a new idea for the VAE-based anomaly detection algorithm. Therefore, this method can only predict short-term anomalies, which has great limitations. However, this method depends on the periodic change law of electromagnetic signals, and the periodicity of electromagnetic signals often changes with time. Then, it judges whether there is an anomaly based on the error value. It uses an LSTM network model to predict IQ channel sampling data of the following four times by learning the past signal sampling values of 32 IQ channels. proposed a periodic anomaly detector that models and predicts IQ channel data. Using the generation characteristics of the VAE, the data can be reconstructed, and the root cause of the anomaly can be analyzed. proposed an anomaly detection method using VAE to reconstruct probability, which is better than the methods based on an autoencoder and a principal component. The two continue to iterate and optimize to achieve the desired effect. The latter reconstructs the timing signal through the generator, and the discriminator judges whether it is an anomaly. The former extracts the potential features of time-series signals by establishing neural networks, reconstructs the signals by features, and distinguishes whether the reconstructed signals are abnormal by evaluating the differences between the reconstructed signals and the original signals. Standard methods include the AE-based method and the GAN-based method. Scholars at home and abroad have proposed many unsupervised learning methods to solve the problems. It has poor real-time performance and cannot be popularized. The traditional anomaly detection model uses complex algorithms and equipment. It has good performance under different SNRs.Īnomaly detection is screening situations contrary to the distribution law of normal data from the data to be detected. Moreover, the model has a simple structure, strong stability, and certain universality. Compared with AE, CNN-AE, LSTM-AE, LSTM-GAN, LSTM-based VAE-GAN, and other models, Accuracy is increased by more than 4%, and it is improved in Precision, Recall, F1-score, and AUC. Experimental results show that the recognition Accuracy of this method can reach more than 85%. Finally, the adaptive threshold is used to determine the anomaly. Then, the normal radar signal data are used for model training, and the mean square error distance is used to calculate the error between the reconstructed data and the original data. Firstly, the signal subsequence is extracted according to the pulse’s rising edge and falling edge. ResNet is used to alleviate the problem of gradient loss and improve the efficiency of the deep network. LSTM is used to discover the time dependence of data. In this method, CNN is used to extract features and learn the potential distribution law of data. Given the low Accuracy of the traditional AE model and the complex network of GAN, an anomaly detection method based on ResNet-AE is proposed. Radar signal anomaly detection is an effective method to detect potential threat targets.
0 kommentar(er)
