A Radar Signal Sorting Algorithm Based on Improved FastICA

The FastICA algorithm is a novel iterative algorithm, also known as the fixed-point algorithm, proposed by Hyvrinen et al, at the University of Helsinki, Finland, unlike conventional neural network algorithms, FastICA uses a batch processing algorithm and requires larger data per iteration. In terms of both distribution and parallelism, the algorithm is still an artificial neural network-based algorithm. fastICA algorithms can be classified into three types: steepness, likelihood maxima, and negative entropy maxima. In this thesis, based on the FastICA algorithm based on negative entropy maximum, the algorithm has faster convergence speed and stronger robustness by introducing Newton iteration algorithm.