Fundamental Compressionist Philosophy.

[Detlef Morgenstern]

Hi Brendan,

--- [Detlef] ---
> > It may be boring, but again: We need not compress the data.
> > We must compress **the theory**!

--- Brendan Macmillan <bren at mail.csse.monash.edu.au> wrote: 
> Is there a distinction?

Point of view (1):
'Data' is what a 'theory' produces.

A numerically controlled (NC) machine tool produces engine blocks.
Compressing those engine blocks you will never distill the NC machine
tool.

If you want to arrive at a compact machine tool you must depart from
a redundant but working machine tool. Not from the engine blocks.

And each step of compacting the machine tool must maintain its full
functionality ('functional compression', 'inductive compression').

The actual point: How can one mechanically transform the engine block
('specification' for the machine tool) into a machine tool prototype
- be it even hopelessly redundant in the beginning.

Point of view (2):
Machine instructions in the main memory of a computer are 'data' for
the one who cannot exploit the 'theory-ness' of those instructions -
their collective ability to (re)produce a given body of data.

He may compact ('zip') them.
No doubt, he may meet the 'less resource consumption' goal.
But he won't distill a 'better theory' - only a ruined one.

I see a distinction as crucial as that I deny the possibility of
distilling a theory from data by simply compressing the data which
stem from the theory.

Detlef


__________________________________________________________________
Do You Yahoo!?
Gesendet von Yahoo! Mail - http://mail.yahoo.de