Copy Link
Add to Bookmark
Report
AIList Digest Volume 4 Issue 138
AIList Digest Wednesday, 4 Jun 1986 Volume 4 : Issue 138
Today's Topics:
Queries - Lisp for Silicon Graphics Machines &
Conditional Independence in Possibility & Curve Fitting Software,
Techniques - Lazy Evaluation,
Psychology - Inside Out,
Description - UMich Cognitive Science and Machine Intelligence Lab
----------------------------------------------------------------------
Date: Tue, 03 Jun 86 10:36:40 -0400
From: ritter@dewey.udel.EDU
Subject: Lisp for Silicon Graphics machines
Our research group (in Chemical Engineering) at the University
of Delaware has just purchased a
Silicon Graphics 3030 workstation, a new version of the 2400 turbo.
We are intersested in obtaining a version of LISP to use with the machine.
Does anyone know what version/company is the best to use?
We plan to build an expert system to predict phase behavior and
are using the Silicon Graphics to display the desired graphics output.
So far, the only company we know of if FranzInc. in CA. which sells
a version of franzlisp.
(We are a little inclined towards a version of Common, since it is
somewhat more universal.)
Any comments or suggestions would be welcomed.
Thank you,
Joe Ritter
ritter@dewey.udel.edu
------------------------------
Date: 3 Jun 86 02:59:23 GMT
From: sdcsvax!caip!seismo!kaist!cskaist!dhkim2@ucbvax.berkeley.edu
(Doohyun Kim)
Subject: Conditional independence in possibility
Hi!
I'm DOOHYUN KIM at KAIST(Korea Advanced Institute of Science and Technology).
It is first time for me to broadcast on these news group.
I'm now search some papers about "Conditional independence of possibility"
These papers are very important in my Master Thesis.
Is there anyone who has following papers ?
Do you have any ideas about how to get these papers?
[1] E. Hisdal, "Conditional possibilities - Independence and
non-interactivity," Fuzzy Sets Systems., vol. 1, pp 283-297, 1978.
[2] E. Hisdal, "A fuzzy 'if then else' relation with guaranteed
correct inference," in Applied System and Cybernetics, G.E. Lasker,
Ed. New York: Pegamon, pp. 2906-2911; also in Fuzzy set and possibility
theory: recent developments, R. R. Yager, Ed. New York: Pegamon, 1982,
pp 204-210.
[3] H. T. Nguyen, "On conditional possibility distributions,"
Fuzzy Sets Systems, Vol 1, pp 299-309, 1978.
Please send a copy to me, if you have.
electlic mail path : dhkim2%cskaist%kaist.csnet@CSNET-REPLY
mail address : DOOHYUN KIM
Dept. of Computer Science,
P.O. BOX 150, CHEONGRYANG,
SEOUL, KOREA 150
------------------------------
Date: Mon 2 Jun 86 12:07:28-PDT
From: Charlie Koo <KOO@su-sushi.arpa>
Subject: Query: curve fitting
I'm interested in getting some information about available software (for IBM
PC) for doing curve fitting. More specifically, given the digitized image
of a circle or part of a circle (2-D), how could we decide:
. whether it is a circle
. what the center and radius of the circle should be?
Thanks.
Charlie
------------------------------
Date: Tue, 3 Jun 86 10:41 EDT
From: Stephen G. Rowley <SGR@SCRC-STONY-BROOK.ARPA>
Subject: Lisp & Lazy Evaluation in AIList Digest V4 #135
Date: Mon, 2 Jun 86 09:27 N
From: DESMEDT%HNYKUN52.BITNET@WISCVM.WISC.EDU
In AIList Digest V4 #134, Mike Maxwell reluctantly prefers the efficiency
of a hand-coded "do" construction in Lisp, although mapping a function on
a list would be more elegant. Indeed, mapping sometimes causes many
unnecessary computations. Consider the following example:
(defun member (element list)
(apply 'or (mapcar #'(lambda (list-element)
(eql element list-element))
list)))
I can't help but point out that, if you're using Common Lisp, this
function is strange on several accounts:
[1] It shadows MEMBER, which can't be a good idea.
[2] It tries to return the first thing in the list eql to element,
whereas the real MEMBER returns a tail of the list or NIL.
[3] It attempts to apply OR, which is a special form and hence cannot be
applied. In this case, you'd use SOME instead.
Obviously, though, I'm just quibbling with your example. Let's move on:
Your statements about the general wastefulness of mapping functions are
true, if you restrict yourself to MAPCAR and friends. However, if you
write your own mapping functions, they can be quite elegant.
Here's an example. I was writing a discrimination net for a pattern
database. Given a pattern, it would search a database for things that
might unify with it, and do something to all of them. (See, for
example, Charniak, Riesbeck, & McDermott's "Artificial Intelligence
Programming", chapters 11 & 14.) For example, a program might want to
print everything that unified with the pattern (foo a ?x), where ?x is a
variable.
The first implementation cried out for lazy evaluation; I didn't want to
compute a list of all the patterns because of consing effects. The
top-level search function returned a stream object (simulation of
laziness) which could be prodded to produce the next answer:
(loop with stream = (search-for-pattern '(foo a ?x))
for next = (next-element stream)
while next
doing (print next))
The second implementation got smarter and made the callers of the search
function package up their intentions in a closure. The search function
would then apply that closure to patterns that it found. The result is
something very mapping-like:
(search-for-pattern '(foo a ?x) #'print)
The second implementation also turned out to be faster and consed less,
although it did use up some more stack space than the first.
Moral: Appropriate use of function closures can often (although not
always) satisfy your needs for lazy evaluation.
------------------------------
Date: Sat, 31 May 86 15:38:33 bst
From: gcj%qmc-ori.uucp@Cs.Ucl.AC.UK
Subject: Inside Out.
This posting is a tangential response to Pat Hayes' posting
in AIList Vol 4 # 125. It is obvious to every child that two
things cannot exist in the same place at once. But a child
does not know what is the other side of the cradle. A child
(and therefore the adult) can never fully expand its spatial
reasoning beyond what the eye can see. Hence, we enter a new
realm; fantasy. For example:-
It is even possible to believe that if I walk into this room,
I will leave reality and enter into a fantasy, eg a film, OR
I wake up one morning and am afraid to open the door in case
I do not recognise the landscape outside.
We carry childhood stories and myths with us to the grave; we
remember the lessons we learnt not only in books and from our
schooling, but also the fairy stories, eg "Alice Through the
Looking Glass" and "The Lion, the Witch and the Wardrobe".
This is more about the distinction between fantasy and reality
than to do with spatial intuition. In the Mind's I, there is
a discussion on whether or not a simulation inside a computer
of a hurricane is any different from the machine's perception
of the real event. To me there would be a world of difference!
In "The Teachings of Don Juan: A Yaqui Way of Knowledge" by
Carlos Castaneda, the author describes, at some point in the
book, his transformation into a bird. His forward begins with
the sentence - "This book is both ethnography and allegory."
But my reading is that he wants you to *believe* his story.
"Choose your own paradigm of reality." -- The Joka.
Gordon Joly
ARPA: gcj%maths.qmc.ac.uk%cs.qmc.ac.uk@cs.ucl.ac.uk
UUCP: ...!seismo!ukc!qmc-ori!gcj
------------------------------
Date: Sat, 31 May 86 15:59:20 bst
From: gcj%qmc-ori.uucp@Cs.Ucl.AC.UK
Subject: Inside Out - Postscript.
The best model we have for the universe is Einstein's theory
of general relativity, which models the cosmos as a 4 dimen-
sional pseudo-Riemannian spacetime. The geometry of such a
model is far from intuitive.
Gordon Joly
ARPA: gcj%maths.qmc.ac.uk%cs.qmc.ac.uk@cs.ucl.ac.uk
UUCP: ...!seismo!ukc!qmc-ori!gcj
------------------------------
Date: Mon 2 Jun 86 10:24:35-PDT
From: Ken Laws <Laws@SRI-AI.ARPA>
Subject: Inside Out
Some thoughts on Pat Hayes' question:
My own intuition is that either 1) the Tardis is simply bigger
inside, as you suggest, or 2) the doorway is a portal to another
dimension or reality that could contain anything at all. One way
to test the intuition is to ask "What would happen if I cut a
hole through the wall?" The answer would not completely distinguish
the two cases, since it is quite possible that the entire wall
(inside and out) is a portal that maps between two realities;
cutting a hole would create a new door with exactly the same properties
as the original, which would tell us nothing. My own guess, though,
is that cutting a hole from outside the Tardis would reveal some
kind of "machinery" or peculiar spatial structure (such as a gel,
crystal, or other "matrix"), while cutting a hole from the inside
would let you out into Gallifrey, Dr. Who's natural environment.
Such portals have been frequent in science fiction, and the exact
properties that we infer for each depends on the author's presentation.
Some conceptions do lead to greater difficulties than others. In
Robots Have No Tails, Louis Padgett (pseudonym) wrote about a box that
was larger inside [partly] because it mapped into the future and the
universe was shrinking. This lead to difficulties at the interface:
things put inside would shrink, but only after a few seconds, and it
is not clear what would happen to a single object such as your hand
that extended across the portal for that length of time. I find
Pat's "shrinking" hypothesis untenable for this reason. Similar
problems arise at the boundary if the doorway is a transporter.
Another such box is the chest that appears in one episode of the
Dungeons and Dragons cartoon on TV. Move it to a particular place,
open it, and you are likely to find a stairway to an alternate
reality. (This is rather like the holes in time used in the Time
Bandits movie.) The D&D chest has the property that the spatial
mapping between realities is fixed and that the portal itself
moves between them. I assume that the box cannot be moved while
open, which helps cover the main conceptual difficulty: why realities
only connect at certain points, and what happens if the box straddles
two such points.
As for something simply being bigger inside, this doesn't bother me.
As we move, we somehow update our internal maps of our surroundings.
As I turn my head, I somehow rotate my mapping of where everything is
relative to my focal direction. It seems unlikely that I actually
store and update a position vector for every book on my bookshelves;
instead I must be storing relative positions of items in the room and
the relative orientation of myself and the room. One could argue that
our natural tendency to build walls, and perhaps even our tendency to
build rectangular rooms, arises from the mental savings in building
these maps in hierarchically partitioned modules with related coordinate
frames. If we store spatial relations in such a manner, it is easy
to see how the spatial relationships inside a box need not be strongly
linked to those outside the box. Just as we can move the box and its
contents as a whole, we can expand it as a whole (on the inside only!)
without affecting our mapping of the rest of the world.
As for two things not occupying the same space, that's not really true.
It all depends on what you mean by "thing". A forest and a tree occupy
overlapping space. Properties such as color and texture certainly
coexist. Fish and streams seem to interoccupy, and groping around
in streams and holes may be a task for which our spatial decoupling
evolved. I wouldn't even be surprised if fish were perceptually
bigger on the inside than on the outside, since they disgorge a lot of
"stuff" when you open them up that is perceived at a different level
of detail than is the smooth exterior of a fish.
-- Ken Laws
------------------------------
Date: Sun, 1 Jun 86 20:06:44 EDT
From: Gary_M._Olson%UMich-MTS.Mailnet@MIT-MULTICS.ARPA
Subject: Description - UMich Cognitive Science and Machine Intelligence Lab
THE COGNITIVE SCIENCE AND MACHINE INTELLIGENCE LABORATORY
The University of Michigan
Ann Arbor, Michigan
The Cognitive Science and Machine Intelligence Laboratory (CSMIL) is an
interdisciplinary organization, spanning the fields of artificial
intelligence, cognitive science, and human-computer interaction. It is
sponsored by three colleges at the University of Michigan: the Graduate
School of Business Administration, the College of Engineering, and the
College of Literature, Science, and the Arts (LSA). Its mission is to
facilitate faculty research and graduate training, with a special focus on
cross-college collaborations.
CSMIL faculty are interested in a variety of specific topics in cognition,
such as vision, attention, learning, reasoning, and problem-solving,
whether they are in humans or machines. Some are also interested in
designing and evaluating the interface between humans and computer systems,
or in developing computer tools to augment and extend human cognition.
CSMIL faculty have a broad range of experience in methods relevant to these
problems, including such areas as the design of special computer
architectures, software design and evaluation, artificial intelligence
programming, and the analysis of human cognition.
CSMIL has a range of specific activities. It sponsors various
seminars, colloquia, conferences, and workshops on the U of M campus in
order to facilitate interdisciplinary intellectual exchange. Many of
these are open to the general technical community in the area as well as to
U of M faculty and students. Periodically, CSMIL will sponsor a single set
of focused intellectual activities designed to stimulate progress in one
particular research frontier, devoting an entire semester or academic year
to intellectual exploration of an important topic. CSMIL coordinates
financial support for faculty projects, and also assists in developing
shared research facilities. CSMIL also takes an active role in
disseminating results from the research of U of M faculty through several
publication series tailored to either specific technical audiences or a
more general readership. Finally, CSMIL has a Corporate Affiliates
Program through which U of M faculty and their peers in corporations can
interact on a regular basis.
For further information about any of these activities, contact:
Gary M. Olson, Director
Cognitive Science and Machine Intelligence Laboratory
The University of Michigan
904 Monroe Street
Ann Arbor, Michigan 48109
313-747-4948
Net address: Gary_Olson%UMich-MTS.Mailnet@MIT-Multics.Arpa
------------------------------
End of AIList Digest
********************
