Using neural networks

[From Shannon Williams (960114.01:00)]

Bill Powers (960114.0915 MST)--

If the brain has the ability to predict how is perceptions will change
if it does and does not emit any behavior, the first project you have to
tackle is explaining how it does this, with a model.

I have ten diagrams now that explain this simply. You can use a trainable
neural network to explain how memory, imagination, language, math, ANYTHING
can work. I even see how the networks could have successively evolved as
species evolved. It's easy.

I have already posted two of the diagrams. You cannot comprehend the
remaining diagrams if you cannot comprehend those two.

Consider our favorite example, a driver steering a car along a road. How
does the driver's brain predict what the path of the car will be if the
driver doesn't emit any behavior?

You can only predict what is predictable. But if my car is veering to the
left, it is easy for me to recognize/predict that it will continue veering
to the left unless I or some obstical does something about it.

Remember that the path of the car is being
affected by bumps and tilts in the road and by variable crosswinds, to
mention just a few of the influences other than the driver's steering
efforts.

Yea so?

Are you visualizing a prediction of one minute or 30 days into the future?
Either visualization would be invalid. What information would the brain
use to predict in these circumstances?

I think that is enough to start with.

Why do you think this? Do you understand how recognition/prediction would
work for a driver? Do you understand how the neural network would be used
to recognize/predict?

-Shannon

<[Bill Leach 960115.22:21 U.S. Eastern Time Zone]

[Shannon Williams (960114.01:00)]

Shannon;

Albus's diagrams "explain everything" too and like most such diagrams it
is impossible to build a working model from the diagram. The diagrams
always contain transforms that do not stand up to rigorous analysis.

... neural network to explain how memory, imagination, language, math,
ANYTHING can work. I even see how the networks could have successively
evolved as species evolved. It's easy.

If it is so damned easy then why doesn't anyone working in this field
provide a working model that will perform to within say .95 of the
behaviour of a real human subject?

You can only predict what is predictable. But if my car is veering to
the left, it is easy for me to recognize/predict that it will continue
veering to the left unless I or some obstical does something about it.

Turning this statement set into a working computer model is, I think, a
bit more difficult than you imagine.

Are you visualizing a prediction of one minute or 30 days into the
future? Either visualization would be invalid. What information would
the brain use to predict in these circumstances?

You bedazzle me! Humans, "predict" for varying periods of time and some
of their predictions are highly accurate even over periods of years!
What are you saying here?

-bill