The estimation of the data model of independent component analysis
is usually performed by formulating an
objective function and then minimizing or maximizing it. Often such a
function is called
a contrast function, but some authors reserve this term for a certain
class of objective functions [36]. Also the terms loss
function or cost function are used.
We shall here use the term contrast function rather loosely, meaning
any function whose optimization enables the
estimation of the independent components.
We shall also restrict ourselves to the estimation of the *noise-free*
ICA model (for noisy ICA, see Section 6), and
assume that the three restrictions in
Section 3.3 that are sufficient for the identifiability of
the model, are imposed.