Re: Computer Learning

From: Bob Tom (tigers@bserv.com)
Date: Tue Apr 04 2000 - 22:55:56 EDT


At 03:51 PM 4/4/00 -0400, you wrote:
> <snip>
> I don't really have any knowledge about the exact workings of the
>computer programming, but speaking from experience at the dragstrip,
>both Bill Tierney and myself have found that (at the dragstrip),
>resetting the computer will result in progressively faster runs. For
>example, if you reset the computer and run a 15.00, then run again,
>it might be a 14.95, then your next run a 14.90, etc. At some point
>this trend stops, but we have noted this behavior time and time again.
>(If you're into bracket racing, you may want to reset your computer
>every time.) Anyway, the only non-WOT throttle operation after the
>computer has been reset involves driving through the staging lanes. In
>a previous post, it was suggested that WOT would halt the learning
>process. Either that isn't true, or the learning that takes place is
>simply from idling through the staging lanes. ? Whatever the case,
>the computer definitely gets faster on each run after being reset.

Maybe the following is also happening (?)

On some 'stock' computers, long term memory cell 15 (the WOT cell)
is being learned at roughly 30-45 mph with a slightly heavy throttle,
just before PE mode is entered. This is about the conditions that happen
when cars return down the access road from the end of the track.

The proposition is that mem loc 15 is being learned each time on the
return road after the run. Also, throughout an entire run down the track,
whatever has been learned in loc 15 is being used to affect fueling.

If the above is true for our computers as well, this means that consistent
fueling
from run to run is not possible as mem loc 15 has potential to learn on the
return
trip unless the compute is reset. Of course, this is just one variable
that we're looking at. I am sure that there are many other variables
that affect performance from one run to another.

Bob

BTW, I have only short term memory and it gets reset every few minutes.



This archive was generated by hypermail 2b29 : Fri Jun 20 2003 - 11:50:36 EDT