topics: |  cognitive | vision | brain

A tour-de-force in cognitively motivated visual computation.  Though Marr
died an untimely death, his legacy continues in the work of Tomaso Poggio and
others.

The most powerful part of the book deals with the early stages of visual
processing. The latter stages, which attributes 3D decomposition models as
the basis for 3D shape recognition, inspired the Geon model by Biederman, but
have not found substantial neurological support yet.

Three levels of information processing

(postulated for any machine performing information processing)


Computational theory        Representation and          Hardware
                            algorithm                   implementation

What is the goal of the     How can this                How can the
computation, why is it      computational theory be     representation
appropriate, and what is    implemented? In             and
the logic of the            particular, what is the     algorithm
strategy by which it can    representation for the      be realized
be carried out?             input and output, and       physically?
                            what is the algorithm
                            for the transformation?


This model has been widely used, e.g. Cosmides & Tooby (1994a, 1995) apply it
to evolutionary biology; “explanations at the level of the computational theory
are called ultimate-level explanations [governed by theory of Natural selection
and derivatives]. Explanations at the level of
representation and algorithm, or at the level of hardware implementations, are
called proximate-level explanations”...

see also: http://www.columbia.edu/~sw2258/materials/PhilPsych/9_Levels.pdf