[From Bill Powers (960929.0900 MDT)]
Avery Andrews (960928) --
Rajesh Rao, in the paper you cited, says this in the abstract you posted (I
would have used quotes from the text, but the file is in postscript, and
with my ghostscript program all I can do is view it on the screen or, I
suppose, copy the file to a disk and take it to Mary's computer which has a
printer -- my laptop doesn't -- and then copy the passages I want for this
post from the printed page by hand. I sometimes truly detest this world of
the GUI).
···
--------------------------------------------------------------------------
Abstract
A characteristic feature of the mammalian visual cortex is the
reciprocity of connections between cortical areas [1]. While
corticocortical feedforward connections have been well studied, the
computational function of the corresponding feedback projections has
remained relatively unclear. We have modelled the visual cortex as a
hierarchical predictor wherein feedback projections carry predictions
for lower areas and feedforward projections carry the difference
between the predictions and the actual internal state. The activities
of model neurons and their synaptic strength are continually adapted
using a hierarchical Kalman filter [2] that minimizes errors in
prediction. The model generalizes several previously proposed
encoding schemes [3,4,5,6,7,8] and allows functional interpretations
of a number of well-known psychophysical and neurophysiological
phenomena [9]. Here, we present simulation results suggesting that the
classical phenomenon of endstopping [10,11] in cortical neurons may be
viewed as an emergent property of the cortex implementing a
hierarchical Kalman filter-like prediction mechanism for efficient
encoding and recognition.
----------------------------------------------------------------------------
What I don't understand is why we need predictors in the visual hierarchy.
Does visual perception concern only the future? When I look around me, am I
seeing the world as it is interpreted to be right now, or as my brain
predicts that it will be some time in the future? I can see that some modest
degree of phase advance would be useful in stabilizing visual control
systems, but is this very elaborate scheme the only way to accomplish that?
Anyway, this paper seems to claim a lot more than it produces. The figures
show the "endstopping" effect as a series of blobs in square frames, and as
some graphs of the results with and without feedback that don't look
dramatically different to me. The author gives his results a VERY generous
interpretation, it seems to me, with the reality of what his model
accomplishes falling far short of what he claims for it.
Maybe you could explain to me what this paper says in terms that help me
understand what's important about it. The functions of the feedforward and
feedback connections in the visual cortex don't seem any clearer to me now
than they did before I read the paper.
Best,
Bill P.