Re: Tha('c ma('c: Fuzzy logic

[Aiviet Nguyen <aiviet@cat.syr.edu>]

Hi Anh Tung,
  Thank you again. I found the communication becomes more exciting.
The fuzzy architecture built on the conventional hardware should be very 
slow. A fuzzy AND as a combination of 4,5 ( or more) conventional operations
will be at least 4,5 times slower. That is why I call it as a simulation.

  About 10 years ago, the first automatic controller made at CERN 
successfully drove a car to a railway station without human intervention.
During the trip, the car should recognise the fuzzy situations and gave 
the best solutions. The fundamental principle based on the idea of neural 
network, which in terms of hardware was implemented by spin glass material.
 The memory about the data pattern was stored in an associated mannner.
Each time the system learned a new thing, it was something like to leave 
a hole on a table. If you have a new input to be processed, it is like to 
drop a ball on the table it will automatically rolls to the nearest hole.
The quite new design principles gave very good performance and real-time 
processing.
Actually, they wrote down something like an action for the system ( the 
system always develops to a minimal action) and found  out that the 
action looked exactly like the Heisenberg model for the spin glass.
And after working on the idea, they could realize all the operations of 
the system by spin glass devices..
  I wonder this result has relations to the recently fashionable quantum
computing that also uutilize the spin materials. I don't have time to 
fill the gap of ten years between twtopics.
  I can only imagine that that would be the real fuzzy hardware. In some 
sense this is the next phase of the pendulum to the analog devices ( of 
course new analog).
  Do you have some ideas or comments?
Aiviet