Connectionists: How the brain works (UNCLASSIFIED)

[Hans du Buf]

On 05/23/2014 06:59 PM, Kelley, Troy D CIV (US) wrote:
> Classification: UNCLASSIFIED
> Caveats: NONE
>
> Hans,
> -------
> Hans wrote:
>
>    
... (deleted) ...
> Anterior insula and dorso-lateral PCC: balance between excessive control and
> lack of control. Imbalance: obsessive-compulsive disorder or schizophrenia.
>    

> -------
> I would be interested in what you think of the very old habituation equation
> first described in 1966 by Thompson and Spencer
>
> We have been able to reproduce this balance/inbalance you speak of above
> between excessive control and lack of control on our robotics systems using
> a habituation equation.
------------------------
Troy,
Please send me (not to the list) a pdf about your simulations.
I cannot find the original TandS paper from 1966 with that magic
formula, probably a few exponentials.
Habituation: no idea how many theories there are by now and back
in 1966 the knowledge was not very complete. Even the Groves and
Thompson "dual process" theory is, well, a theory. Thompson himself,
in his "history" paper, wrote that "both short- and long-term
habituation are far more complex than earlier believed." I bet it is.
In my earlier email I wrote about corollary discharges for warning
sensory areas about upcoming (ego)motor actions. Habituation is related
to external events. A system that is continuously predicting reward
and attracted by surprise (entropy) must shield itself from distractions
which do not yield reward in a broad sense - both sensory and motor
processes cost energy. In contrast to snapping at a baby in a cradle,
Pavlov's dogs were rewarded! But, temporal conditioning is also habituation.
Reinforcement learning may be related, but the crucial difference is
reward vs. distraction, learning useful things and (learning to)
suppress useless things. In a famous learning task (vernier acuity)
subjects were trained during days, and the effect lasted then for
months or even half a year. So what is the reward here? Subjects won't
be able to read better because this is line acuity. And if you don't tell
that they will be asked to repeat the experiments half a year later,
why remember?
Hornet's nests are fascinating...
Hans

-----------------------

>
>
> I would be interested in knowing what you think the habituation equation
> DOESN'T capture.  My sense is that it doesn't capture reinforcement learning
> very well.
>
> Thompson, R. F.,&  Spencer, W. A. (1966). Habituation: a model phenomenon
> for the study of neuronal substrates of behavior. Psychological review,
> 73(1), 16.
>
> Note especially the nine characteristics of the equation they outline nicely
> beginning on page 19 of the article.
>
> Troy Kelley
> Cognitive Robotics Team Leader
> Human Research and Engineering Directorate
> Army Research Laboratory
> Aberdeen, MD, 21005
> V: 410-278-5869
>
>
>    
-- 

=======================================================================
Prof.dr.ir. J.M.H. du Buf                          mailto:dubuf at ualg.pt
Dept. of Electronics and Computer Science - FCT,
University of Algarve,                            fax (+351) 289 818560
Campus de Gambelas, 8000 Faro, Portugal. tel (+351) 289 800900 ext 7761
=======================================================================
UALG Vision Laboratory:            http://w3.ualg.pt/~dubuf/vision.html
=======================================================================