Just to verify that I am reading you, the additive classifier *should*
do a horrible job exactly because N+1 << 2^N. Selection is nullified by
linearity, while modularity benefits from linearity. As a consequence,
and through an evo-devo lens, biology optimizes about this paradox.
Do I have this right about selection? In a recent conversation with Nick,
I got the impression that this paradox of heritability is the source of
one of his *bugs*.

"""
A much better decomposition, of course, is not to use structureless
sets like cumulants, but rather to find representations of algorithmic
architectures that can be put into a frame of statistical identification.
"""

Are cumulants structureless? I keep thinking of them as being analogous
to moments and effectively differential information, but OTOH, maybe all
the *logs* stills this. Idk, I am a newbie here. The other idea your
comment calls to mind is the work being done to *learn algebraic
varieties*. Like varieties (the geometric objects they are) are the
consequence of an underlying algebra of operations (polynomial rings),
representations of programs (as denotation-level objects) can be learned
despite the inevitably wide variety of implementations.

Thank you for the references and the consideration.
-. --- - / ...- .- .-.. .. -.. / -- --- .-. ... . / -.-. --- -.. .
FRIAM Applied Complexity Group listserv
Fridays 9a-12p Friday St. Johns Cafe   /   Thursdays 9a-12p Zoom  
bit.ly/virtualfriam
un/subscribe http://redfish.com/mailman/listinfo/friam_redfish.com
FRIAM-COMIC http://friam-comic.blogspot.com/
archives:  5/2017 thru present https://redfish.com/pipermail/friam_redfish.com/
  1/2003 thru 6/2021  http://friam.383.s1.nabble.com/

Reply via email to