A detection method and device for an input signal of an unknown complex system

Abstract

The invention discloses a method and a device for detecting an input signal of an unknown complex system. The detection method of the input signal of the unknown complex system includes: obtaining the observable output signal of the unknown complex system; constructing a high-dimensional observation space according to the observable output signal; constructing the low-dimensional observation space of the unknown complex system according to the high-dimensional observation space The linear evolution trajectory and the equivalent input driving signal in the three-dimensional approximate phase space; the input signal of the unknown complex system is detected according to the linear evolution trajectory and the equivalent input driving signal, thereby effectively improving the input signal of the unknown complex system detection reliability.

Description

technical field [0001] The invention relates to the field of information technology, in particular to a method and device for detecting an input signal of an unknown complex system. Background technique [0002] Reliable data processing and information extraction in complex and unknown mechanism scenarios play an increasingly critical role in exploring basic natural sciences (such as the discovery of biological and physical mechanisms) and engineering practical applications (such as information transmission in complex scenarios). In this context, the best signal detection and information processing methods represented by Bayesian statistical methods will face great challenges. In many practical scenarios, due to the high complexity (such as involving complex linear dynamics or nonlinear dynamics) and unobservability of the unknown system to be analyzed, the unknown system can only be detected or inferred by means of the observable system output response Enter information. ...

AI Technical Summary

Problems solved by technology

However, at present, such model-free methods still mainly rely on a large number of training samples (huge training-corpus) and a strong supervised learning process (strong supervised-learning), and it becomes difficult to perform in unlabeled, small-sample signal detection and information processing scenarios.

In addition, most unknown complex systems to be analyzed exhibit dynamic characteristics, that is, given the same input, the internal mechanism and observable output of the system also change dynamically over time. The learning effect is not good, and the dynamic migration characteristics of the system cannot be tracked in time, resulting in poor performance, and even the learning process diverges, failing to achieve good results

Images