[ACT-R-users] Efficient approximation of the Base-Level Learning Equation

[Lynne Reder]

Alex, two things:

  first in my step-child version of ACT-R's non-procedural memory  
(called SAC), we use a transient boost in current activation that  
decays exponentially.  I think old versions of ACT did that too, but I  
know longer remember.  In any case,  I've been using a current boost  
that decays exponentially for about a decade or so (first published in  
1996).  Second, as you may know, Barbara Dosher & co. have evidence  
that power law learning is really an artifact of averaging different  
exponentials from different processes within a person or across people.  
  I heard her give a talk recently.  Perhaps I have mischaracterized  
these things, but thought you might know about this and want to address  
that.

Regards,
Lynne


On Feb 15, 2006, at 4:19 PM, Alex Petrov wrote:

> Dear ACToRs,
>
> As you know, the base-level equation is one of the pillars of ACT-R's  
> success.
> It has a serious practical drawback, however -- it consumes a lot of  
> memory
> and CPU cycles.  The approximate formula published in the 1998 book is  
> good
> for many purposes but does not capture the transient boost after each  
> use of
> a chunk. I have developed an improved approximation, which does take  
> all
> critical properties of the exact equation into account. It was used  
> with
> great success in the ANCHOR model of Petrov & Anderson (2005).  It can  
> be
> very useful for large simulations, allowing ACT-R to scale up to more
> realistic memory sizes and more prolonged learning periods.  As far as  
> I know
> (though I'm not sure), it is already implemented in the current  
> version of
> ACT-R.  (Dan, can you please advise on that?)
>
> A short two-page description of this new approximation has just been  
> submitted
> to ICCM06.  It is attached below, together with its associated MATLAB  
> code.
> All this stuff will soon be posted on my web page at   
> http://alexpetrov.com
>
> I do hope this will be helpful to the ACT-R community.
>
> Best regards,
>
>   -- Alex
>
> -------------------------------------------------------------
> Alexander A. Petrov:  apetrov at alexpetrov.com
>
> Post-doctoral Researcher
> Department of Psychology
> University of Colorado, Boulder
> http://alexpetrov.com
>
> It is better to light one candle than to curse the darkness.
> -------------------------------------------------------------
> <ApproxBaseLevelEq.pdf><ApproxBaseLevelEq- 
> MatlabCode.zip>_______________________________________________
> ACT-R-users mailing list
> ACT-R-users at act-r.psy.cmu.edu
> http://act-r.psy.cmu.edu/mailman/listinfo/act-r-users
>


========================================================

Lynne M. Reder
Professor
Carnegie Mellon University
Pittsburgh, PA 15213
412-268-3792 (office)
412-268-2844 (fax)
http://www.andrew.cmu.edu/~reder/reder.html  (home page)
reder at cmu.edu  (email)
-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: text/enriched
Size: 2813 bytes
Desc: not available
URL: <http://mailman.srv.cs.cmu.edu/pipermail/act-r-users/attachments/20060215/09fe7fb1/attachment.bin>