Table 1 Nonlinearities g used in FastICA. The nonlinearities probe the non-Gaussianity of the estimated components in different ways.
From: Statistical model of natural stimuli predicts edge-like pooling of spatial frequency channels in V2
g1 (y) = tanh (y) | Classic measure of sparseness |
g2 (y) = y exp(-y2/2) | More robust variant of g1 |
g3 (y) = y2 | Skewness (asymmetry) |
g4 (y) = exp(-y2/2) | Robust variant of g3 |