=> PAT

PatternActivatedTransformations


(pat's not a lang or a framework; what is pat?;)
someday-objective:
to be an (reference)example of[an]new[age]lang/concepts
that should grow, w/ref to other langs/AI techniques
ok that was wishy-washy
I want to have a mop/clos-KR/reasoning-extension
that spans frame&DL &can easily be used
for the class of problems I'm interested in.
[There are separate systems but they
do not span&aren't(MOPd)in(to)the ObjSys]

=> plan

I plan to work on a reference implimentation of this idea;
that would be made-of/like (the CLOS based part of) lisaon top of franz'sacache /prolog/etc
w/the hope of giving a bit more
=> I hope to get a persisitent, fwd-chainer on the prolog/etc qry system
but have it more flexible/widely-useful; mixing in some TMS/MBR ideas.
I'd also like to mv(all)the code in w/the ontopersisting/qry/etc it all.
(In fact maybe a ModelBased view of the different lib/functionality I want
that would include the ability to gracefully still work if some libs fail.
I think this is more easily done w/the idea of models of different depth.)

=> Maintainer is Mike Bobak started Sept9,2006.

I don't take (lang/framewrk/etc)suggestions from any1, unless they provide @least some top/hi levelPcode
by Pcode (i mean pseudo-code;)

=>IPwise

I tend toward wanting a GPL/BSD sort of distribution of code, esp. as I hope to mainly use other's code
but I will reserve the right to decide on the best way to share my contributions; as they become real/(more than an idea)

=> ps. for the protege/etc. community

I want to end the dependence on jess/swrl and java/xml-on the backend in general
I want to make the whole biz/etc-intel /KA/KnMngt thing easier;(no bothering w/many pat rules;)
I might try algernon w/the real-lisp~sys in the(net)background; to see how/(if)it is worth linking it all(as > a gui).
host?
08:(all updates to the top) I'm about to get a new mac;where I will revisit several large bits of software, some of which will benefit from the new platform; One will be using HDF for graph storage, as it is already well used, &I only see more gobs of data coming online, not all of which will be shoved into a triple-store; There are already gobs of (yes even realtime-event data)being shoved into this format;&it shouldn't be terribly difficult to use(to some degree);It already has lots of parallel-io/grid enable software, which will also be very useful, in this cloud/oriented world. ;It could also keep me from being stuck w/the commercial allegro-graph(which if it is used commercially has a terrible contract).

I have to start(aug07)putting updates near the top:.. layered kn-base, small(neutral)machinary rest in a well-organized KB
........*..
KnowOS/BioBike.....**;:...
more biobike:
Jeff Shrager posted links to his ILC 2005 paper, “KnowOS: The (Re)Birth of the Knowledge Operating System” along with slides and audio from his talk.[Deductive Biocomputing]

pps. have thought about this.

I'll have to add more (related)math/stat-s/sim/etc here soon.

Dec(06)update:
I've been having more DRY thoughts, and some other things that have been bothering me which were nicely summed up in a talk I just caught; It has inspired me; to be a bit more daring by enacting some of the many notes that never make it to this page.

2007 udpate: Might try to go w/SBCL? Also to get a owl/DL /allegroGraph flavor, I had found that w/clips/protege at least that you can have class restrictions on slots, to get something like a hierarchy of relations.

Now I'm starting to thing of a twist on a (garnet like) proto-ins system, where all the interesting sets would represent all the sn:val (and even relationships)once if possible; and might be able to be learned from incoming data.

Might be able to get the PatternActivatedTransitions build up these(higher level)descriptions. Learned from data (where from any view, there would be a prot/set describing that aspect/view); Also where the transition/rules could be used to make commentary that wasn't (necessarily)in he orininal col/sn:vals; Where SME-Kn can use any of these data(sets)descriptions to create a higher-level description(s) [that can be re-cursed/applied] to get closer to the problem description closest to the mapping to a suggested answer to your problem.

I will still try to look at Algernon, Loom, and OKBC for possbile use/ideas.

[I'd still love to play w/allegro-cache(&graph or KnowledgeWorks, if I could); and plan to if someone else buys it; but I couldn't easily justify the 4k for KnWrks or who knows how much if I wanted to sell a franz app]_So SBCL will also be my friend in these endevours.[Openmcl is also possible (it got me back in w/GL on the mac very quickly;)] I'm hoping for something that will aid in simply scalable concurrent computation.

[Early Spring]-I'd really like to get back to this, &some related interoperability issues.

I'm getting more familiar with the Allegro products, but also realize I can't depend on them being there, especially if I hope to open-source this; So I hope to use SBCL (&maybe even some cusp based dev, w/some protege too).

ps. I saw Alan_Kay speak* in the bay, just before I left last time. It really resonated w/past kiss/dry desires, and pushes me to take from present systems rather than just composing them (w/great chance of code bloat/etc).

07updates:
I would like to have a triple-store that worked well with massive amounts of numerical data, and let the semantic net retain as much of the clos/obj/ins feel as possible. So somewhat influenced by dro-device, and maybe having HDF/ postgreswProlog underneath for persistence/ease of interaction and hopefully very quick reactive/deductive reasoning over huge amounts of data (if well designed). Hopefully w/(clustered)/concurrent, maybe cl-muproc/ star-Lisp like ability. I want to deal w/much larger chuncks of data (natively) w/o having to have it all in memory. Maybe somewhat like Allegro(Cache/Graph), but w/an eye to larger data, and almost no worry about xml until absolutely necessary. I'm ok w/interop w/java/etc, but that should fall out easily and not alter a simple/clean design.

I want a new(protege)tab: like jess-tab(w/o jess)& okbc/algernon(w/o the full api/system)
Having a cusp connection could be nice too(vs just a jesstab like view).
=Most of this is to have a consistent interface, and some KB/KM support, but if I could do it all in Lisp I would.
I'm considering a web based interface, if it could be as smooth.
So maybe the Protege+Lisa(picture)might be more in spirit than code.

Still thinking of something between frames and full owl/dl

Mid 07:1st post on new bioinfo page seams appropos.

summer: Still like dro-device, ideas; to get a CLOS like view on OWL like semantic(nets); Could use Protege frames view; to ediit owl-like(extra constraints); while connected to a Lisp, using say lisa.sf.net &other reasoning as needed.

fall:still acl-cache/graph or..; might be nice to play off postgres* rule system
I would really like a KR/reasoning system like the QR work that I've seen, but as a MOP extension to CLOS.
Between closed&open-world-assumptions; I like to think one a subset of the other; If you can't get a y/n, it just depends on what you do w/maybe. If I haven't said it before, I always thought a MBR/CompositionalModeling/cml approach would be taken to semantic-web-service composition.+

Dec(07)note:

I'd like an oop/model(ing) situation, that give action to ontology
 where a symbol is never lost from a namespace, it is always found by nearest distance
 if that isn't enough, what is given/expected as a signature to match, as I'm also hoping
 that several methods could be used for the same verb, given context /time, needs..

Dec(09)

.. _ . 

.  . .  . 
 
. . . 
. . 
. 
. . .  . . .
2010
.
Also like scone and still think of aGraph like interface
but I might think of wrapping really big data w/in HDF
Consider one generic- interface that could use any/all
w/a MOP(like?) view into it.