From newshub.ccs.yorku.ca!ists!helios.physics.utoronto.ca!news-server.ecf!utgpu!cs.utexas.edu!uunet!trwacs!erwin Tue Jun  9 10:06:29 EDT 1992
Article 6045 of comp.ai.philosophy:
Path: newshub.ccs.yorku.ca!ists!helios.physics.utoronto.ca!news-server.ecf!utgpu!cs.utexas.edu!uunet!trwacs!erwin
>From: erwin@trwacs.fp.trw.com (Harry Erwin)
Newsgroups: comp.ai.philosophy
Subject: Grounding
Message-ID: <617@trwacs.fp.trw.com>
Date: 2 Jun 92 17:13:11 GMT
Organization: TRW Systems Division, Fairfax VA
Lines: 29

In scanning the literature, I found a most interesting paper in NIPS 2. In
"Reading a Neural Code," Bialek, Rieke, de Ruyter van Steveninck, and
Warland (NIPS 2, 36-43, 1990) point out that in the neuron they were
studying, there is insufficient time to integrate the firing rate to
generate an estimate of the signal strength. Instead, the organism has to
decode a short segment of a spike train. They discover that this can be
done with a simple linear filter.

The point of this paper is that the spike train _encodes_ a continuously
varying stimulus, and the decoding is relatively simple. A constant
stimulus strength has to be converted into a periodic function to be
encoded and passed back into the brain. This is generally done by a motor
process. Hence the qualia we process are not simple measurements, but are
transforms of the external data into a frequency domain. Now the reverse
transforms do occur, resulting in spatial maps, but even then the
frequency data are retained through much of the processing. (See "Neuronal
Maps for Sensory-Motor Control in the Barn Owl," Spence, Pearson, Gelfand,
Peterson, and Sullivan, NIPS 1, 366-374, 1989, for an interesting example
where frequency spectra are retained to tag specific patterns during
processing.)

Hence my perceptions of an object have been transformed into the frequency
domain and back at least once.

Cheers,
-- 
Harry Erwin
Internet: erwin@trwacs.fp.trw.com



