ABSTRACT
This chapter introduces the well-known concept of linear neural network, in particular, the one made up of one neuron. The minor component analysis (MCA) is a technique extracting the second-order statistics of the input signal and plays an ever increasingly important role in data analysis and signal processing. MCA has several applications, especially in adaptive signal processing. It has been applied to frequency estimation, bearing, beamforming, moving target detection, and clutter cancelation. The analysis of the temporal behavior of all MCA neurons can be carried out by using the stochastic discrete laws, since the mere use of the ordinary differential equations approximation fails to reveal some important features of these neurons. The MCA learning laws are instantaneous adaptive gradient algorithms and then work sequentially. If incoming inputs are collected in blocks and are fed to the neuron, which changes its weights only after the whole block presentation, all methods typical of the batch learning can be used.
