Abnormity detection method based on BQP network

Abstract

The invention belongs to the technical field of deep learning, and particularly relates to an abnormity detection method based on a BQP network, comprising the following steps: S1, presetting an abnormity detection image training data set; S2, building a BQP network, and setting parameters; S3, for each batch of training batches fed into the BQP network, extracting features in the image by utilizing a feature extraction network in the BQP network, and outputting a batch feature vector X with the size of B*n; S4, constructing a characteristic hypersphere in a QP output layer in the BQP network,wherein the QP output layer outputs an optimal dual variable; s5, through the classification loss function and the consistency loss function, the loss function is calculated for the feature vector Xoutput by the feature extraction network and the optimal dual variable of the QP output layer respectively, and parameter optimization is carried out on the BQP network through a back propagation algorithm; and S6, during detection, comparing the feature vector modulus length output by the feature extraction network with a set threshold value to realize abnormal detection.

Description

technical field [0001] The invention belongs to the technical field of deep learning, and in particular relates to an abnormality detection method based on a BQP network. Background technique [0002] With the explosive growth of data in recent years, people's demand for intelligent analysis of data is getting higher and higher. Among them, anomaly detection technology, as a branch of machine learning, has played an important role in intelligent analysis of data. For example, in massive video surveillance data, through computer anomaly detection technology, it is possible to more accurately locate the time period of abnormal events, which greatly reduces labor costs. [0003] Anomaly detection is the determination of samples other than well-defined or identified normal samples, which is a special detection problem. Anomaly detection involves machine learning, data mining, mathematical statistics, information theory and other related knowledge, widely used in intrusion detec...

AI Technical Summary

Problems solved by technology

[0004] The complexity and difficulty of anomaly detection lie in the following aspects: First, different from traditional classification problems, in anomaly detection problems, the number of samples of normal events is far more than the number of samples of abnormal events, and the difference between normal and abnormal training data is There is a serious imbalance in the number of

Secondly, the number of abnormal samples in a training set is limited, but it may contain multiple types. At the same time, the abnormal samples in this training set still cannot exhaustively contain all abnormal situations, so it is difficult to find a suitable model to describe The data

Moreover, due to the noise of the data itself or the noise introduced by the system, it is often difficult to identify noise and anomalies, which brings a lot of uncertainty to the anomaly detection problem.

Images