INTRODUCTION TO THE TLS EXIN NEURON - Neural-based Orthogonal Data Fitting: The EXIN Neural Networks - page 130

Figure 4.6 κ = 0 . 4; averaged over 80 independent trials for w 5 .

Figure 4.7 κ = 0 . 4; averaged over 80 independent trials for the index parameter (expressed

in decibels).

2 increases
over 0 . 2, TLS GAO diverges. This phenomenon can be explained considering that
because of its learning law linearization during the transient and in presence of
the high noise, the neuron weights are so far from the solution to find themselves

slightly less than 15 dB with respect to TLS GAO. Interestingly, as σ