Regularity

[Toews, Robert]

FYI:  The visual effect Gerry is referring to can be easily demonstrated by
the Cornsweet illusion which is a sequence of light to dark bars, each bar
uniformly shaded within itself but slighly darker or lighter than those on
each side respectively.  The bars will appear to be non-uniformly shaded
such that the portions near the edges enhance the contrast.

I'm not sure I follow the application to CasC though.  The statement Gerry
makes about coding edges requiring less information than coding every pixel
sounds like the argument for using vector scan rather than raster scan in
graphics displays.  When the information to be displayed was simple, like
wire drawings, it was more efficient to use vector scan.  However, now that
displays have become much more complex, it is more efficient to use raster
scan, and vector scan is obsolete.

Bob

> -----Original Message-----
> From:	Gerry Wolff [SMTP:gerry at sees.bangor.ac.uk]
> Sent:	Tuesday, February 02, 1999 12:33 PM
> To:	CasC Maillist
> Subject:	Re: Regularity
> 
> Sergio Navega wrote:
> > 
> > As usual, Gerry Wolff's message is both insightful and
> > thought-provoking.
> 
> Thanks for those nice words, Sergio!
> 
> ...
> 
> > Back to the beginning:
> > 
> > > Gerry Wolff wrote:
> > >[snip]
> > >Well, this is just one simple example. What about z = x + y (closer to
> > >your example)? The table would look something like this:
> > >
> > >x y z
> > >1 1 2
> > >1 2 3
> > >2 1 3
> > >1 3 4
> > >3 1 4
> > >etc (very tedious!)
> > >
> [Text deleted: please refer back]
> > 
> > This is in the eye of the tornado! I guess we can find several other
> > examples with a similar outcome: repetition.
> > 
> > What I think is missing in this scheme to boot it up is something that
> > stands for *learning*. I mean for learning, in this context, the ability
> > of *improved performance* as the number of experiences of the agent
> > grows.
> > 
> > To exemplify what I mean, I'll resort to a practical example.
> > Suppose we want to design one agent capable of interpreting simple
> > images. How can we teach this agent to recognize a tree? I'll
> > reduce the problem here to fit the space of this message (also
> > because I have to workout on other details of the problem :-)
> > 
> > Imagine that the following pattern is a symbolic representation of
> > the digitization of the edge of an image of a trunk of a tree
> > (taken as a single horizontal scan):
> > 
> > a
> > aaa
> > aa
> > aaaa
> > a
> > aaa
> > aa
> > aaaa
> > a
> > aaa
> > aa
> > aaaa
> > 
> 
> [Text deleted: please refer to Sergio's original message for the full
> text.]
> 
> I like your examples and suggestions about possible ways forward.
> 
> Regarding 'edges', I thought I might mention again something I wrote
> about in "Computing, Cognition and Information Compression" and which is
> probably quite familiar to people with a psychology or neurophysiology
> background. For people who may not have come across these ideas, it has
> been known for some time that in the eyes of most kinds of animal there
> is a 'lateral inhibition' mechanism which has the effect of picking out
> some kinds of edges (eg between a uniformly white area and a uniformly
> black area).
> 
> In the black area, the receptor cells fire at a 'background' rate. In
> the white area, they are stimulated to fire at a higher rate. But there
> are inhibitory fibres from each receptor to the surrounding cells, and
> each cell receives inhibitory fibres from its neighbours. So the cells
> in the white area inhibit each other and their rate of firing is pushed
> back to something close to the background rate. This is true for all
> parts of the white area except the edges, where relatively few
> inhibitory signals are being received from the black area. So, around
> the edges, the rate of firing is relatively high (and this produces an
> inhibitory trough around the black area). The overall effect, is to mark
> the edge between black and white! Neat isn't it?
> 
> This mechanism can be interpreted in terms of information compression
> (coding the edges requires less information than coding every 'pixel'
> everywhere) and it can also be interpreted in terms of repetition.
> 
> That's all for now.
> 
> Gerry