Though great advances in physiology and microbiology have in recent
years illuminated the cellular apparatus of neural codes, the design
of neural systems and the evolutionary reasons for this design are
less well understood. This talk will describe one approach to the
question of why the human visual system is structured as it is,
namely the argument from efficiency. Statistical predictabilities in
the visual environment place restrictions on the sorts of neural
codes that evolution would select as efficient, though the idea of
biological efficiency must be carefully defined in this context. I
will give an example of how retinal codes appear to be matched to
regular statistics of the natural world. Then I will extend this line
of reasoning to human-made art, where statistical regularities can
grant insights into both vision coding and art-making. The
statistical properties of art that are relevant to the neural coding
of vision are also of potential use for developing a "stylometry" of
artworks. Stylometric techniques under development could be useful
for performing authentication and historical ordering and for
predicting viewer preferences.
|
