Recursive Kernel Adaptive Filtering Method Based on Kernel Function

Abstract

The invention discloses a recursive kernel adaptive filtering algorithm based on kernel function, which includes S1 initializing variance parameters, kernel parameters, order parameters and network parameters; The function is mapped to the high-dimensional feature space; S3 multiplies the mapping result of the input signal in the high-dimensional feature space with the weight in the high-dimensional feature space to obtain the output result of the current iteration, and uses the output result and the expected output in the high-dimensional feature space The deviation value in the space; S4 inputs the input signal, the mapping result, the output result, the expected output and the deviation value into the neural network, and trains the neural network; S5 judges whether all groups of training data have been used, if so, enter step S6, otherwise Select a set of unused training data, accumulate i once and return to step S2; S6 uses the trained kernel adaptive filtering model to filter the new input signal.

Description

technical field [0001] The invention relates to filtering technology, in particular to a kernel function-based recursive kernel adaptive filtering method. Background technique [0002] The kernel adaptive filtering algorithm is widely used in the current hot machine learning related fields; the basic idea of ​​the kernel adaptive filtering algorithm is to establish a model, map the input data to the feature space through the kernel equation for processing, and process the result Modify the parameters of the model itself to get an output closer to the expected value, perform this process iteratively, input new data for training, until the processing result reaches the expected value, the model can be used to solve problems such as time series prediction, pattern recognition and classification , Dimensionality reduction, image processing, data compression and reconstruction, nonlinear regression and other issues have important practical significance. [0003] Currently common...

AI Technical Summary

Problems solved by technology

[0005] In view of the above-mentioned deficiencies in the prior art, the recursive kernel adaptive filtering method based on the kernel function provided by the present invention solves the problem of large errors in the existing KLMS and KRLS kernel adaptive filtering

Images