distributive NARS

[YKY (Yan King Yin, 甄景贤)]

Dear NARS,

How's it going?  Long time no see! =)

I've proposed a distributive architecture that is generic enough to fit many AGI inference engines, including NARS, Opencog's PLN, and Genifer.  Lately I have been re-factoring the inference algorithm, thus gaining some insight that allowed me to generalize it.

The explanation on the blog is brief, so feel free to ask questions about it.  Basically, NARS can retain its uncertain calculus independent of other engines.  I've forgotten how NARS deal with unification (as in predicate logic), even though this aspect can be factored out too.  So there is a lot of flexibility.

Regards,

-- 

KY

Notice:  50% of all my income derived from AGI will be instantaneously donated to charity.

"The ultimate goal of mathematics is to eliminate any need for intelligent thought" -- Alfred North Whitehead

[Pei Wang]

2011/7/6 YKY (Yan King Yin, 甄景贤) <generic.in...@gmail.com>:

> Dear NARS,
> How's it going?  Long time no see! =)

I'm on a vacation/sabbatical, working on a new book on NAL. After that
I will try to finish a new version of NARS.

I'll attend AGI-11 next month. For the other recent events I'm
involved in, see http://www.cis.temple.edu/~pwang/

> I've proposed a distributive architecture that is generic enough to fit many
> AGI inference engines, including NARS, Opencog's PLN, and Genifer.  Lately I
> have been re-factoring the inference algorithm, thus gaining some insight
> that allowed me to generalize it.

I cannot access blogspot from China. You'll have to use email.

NARS is easy to be turned into distributed, because each "concept" in
it is a natural unit for processing and storage, and can work
independent of the other concepts.

However, the inference rules of NARS and its memory/control are
closely coupled, and the latter is very different from those of the
other two you mentioned. I'm not sure how far you can go in
"generalizing" them into a common (and no-trivial) architecture.

> The explanation on the blog is brief, so feel free to ask questions about
> it.  Basically, NARS can retain its uncertain calculus independent of other
> engines.  I've forgotten how NARS deal with unification (as in predicate
> logic), even though this aspect can be factored out too.  So there is a lot
> of flexibility.

Unification in NARS is very similar to that in logic programming
(e.g., Prolog). The source code is at
http://code.google.com/p/open-nars/source/browse/trunk/nars/language/Variable.java
(the last several methods).

Pei

> Regards,
> --
> KY
> Notice:  50% of all my income derived from AGI will be instantaneously
> donated to charity.
> "The ultimate goal of mathematics is to eliminate any need for intelligent
> thought" -- Alfred North Whitehead
>

> --
> You received this message because you are subscribed to the Google Groups
> "open-nars" group.
> To post to this group, send email to open...@googlegroups.com.
> To unsubscribe from this group, send email to
> open-nars+...@googlegroups.com.
> For more options, visit this group at
> http://groups.google.com/group/open-nars?hl=en.
>

[YKY (Yan King Yin, 甄景贤)]

On Thu, Jul 7, 2011 at 7:59 AM, Pei Wang <mail.p...@gmail.com> wrote:

 

I'm on a vacation/sabbatical, working on a new book on NAL. After that
I will try to finish a new version of NARS.

I'll attend AGI-11 next month. For the other recent events I'm
involved in, see http://www.cis.temple.edu/~pwang/

The AGI-11 workshop on self-programming sounds really nice, but I won't be able to attend.  In general I'll be staying in Hong Kong and won't go to conferences unless they're very close (to be environmentally friendly).  But I'm planning to visit ZhongShan Univ some time.

I cannot access blogspot from China. You'll have to use email.

It's OK, I'll write it in my PDF book and send it to you later...

NARS is easy to be turned into distributed, because each "concept" in
it is a natural unit for processing and storage, and can work
independent of the other concepts.

Right.

 

However, the inference rules of NARS and its memory/control are
closely coupled, and the latter is very different from those of the
other two you mentioned. I'm not sure how far you can go in
"generalizing" them into a common (and no-trivial) architecture.

If you mean inference control, that is something I have not looked into.  My distributive architecture only tries to answer queries.  Any optimization has yet to be done at the distributive level, though I have some ideas of how to do that.

Unification in NARS is very similar to that in logic programming

(e.g., Prolog). The source code is at
http://code.google.com/p/open-nars/source/browse/trunk/nars/language/Variable.java
(the last several methods).

That sounds good, the NARS can retain its own unification or matching methods, my distributive architecture does not fix that.

I'll send you a more detailed description in a week or so...

Thanks!

KY