Jesus Malo, 9/26/2001
<br>
<br>
Inversion of the divisive normalization revisited
<br>
<br>
Last June I reported a fast technique to invert the divisive
normalization. However, fundamental problems have appeared in the most
interesting cases (using optimal parameters in the normalization). I
will discuss:
<br>
More specific conditions to ensure the invertibility of the
normalization.
<br>
Conditions for component independization using the divisive
normalization.  Can we expect true component independization (PDF
factorization) with an optimized normalization?
<br>
Maximum likelihood versus optimal prediction in the estimation of
parameters of the divisive normalization