clarke@csri.toronto.edu (Jim Clarke) (11/02/88)
(GB = Galbraith Building, 35 St. George Street)
(SF = Sandford Fleming Building, 10 King's College Road)
SUMMARY:
GRAPHICS SEMINAR - Tuesday, November 15, 3.00 p.m. in GB221 -- Bruce F. Naylor:
"Binary Space Partitioning Trees -- A New Representation of Polyhedra"
AI SEMINAR - Thursday, November 17, 11.00 a.m. in SF1105 -- Se June Hong:
"Knowledge Engineering Experience Gained from a Large Financial
Marketing Expert System, FAME"
-------------------
GRAPHICS SEMINAR - Tuesday, November 15, 3.00 p.m. in GB221
Bruce F. Naylor
AT & T Bell Laboratories, NJ
"Binary Space Partitioning Trees
A New Representation of Polyhedra"
Abstract. Geometric modeling is concerned with the problem of
representing geometric objects and performing operations on these
objects efficiently. We will show that Binary Space Partitioning
Trees (BSP Trees) can represent polyhedra of any dimension and
that there are simple efficient algorithms that operate on them.
The BSP tree is a binary tree that represents a recursive parti-
tioning of space by hyperplanes (in 3D, planes). Hyperplane
equations are stored at internal nodes while leaves correspond to
unpartitioned convex regions of space.
Operations on BSP trees include:
Affine transformations : rotation, scaling, translation, etc.
Set operations : union, intersection, difference, etc.
Visible surface determination
Ray-tracing
Volume, surface area, moments
Conversion to and from boundary representations
Each operation is relatively simple and all use the same basic
techniques. A number of these are demonstrated in a video that
will be shown in which a user interactively designs solid objects
by sculpting. This is the first instance of such capability,
achieved with application code running on a commerically avail-
able graphics workstation.
AI SEMINAR - Thursday, November 17, 11.00 a.m. in SF1105
Se June Hong
IBM T.J. Watson Research Centre, NY
"Knowledge Engineering Experience Gained from a Large Financial
Marketing Expert System, FAME"
Abstract. In this talk a complex domain, that of financial mark-
eting, is described, and our use of this domain to advance metho-
dologies for large knowledge based systems is presented. We have
developed a framework based on an object-centered knowledge
representation paradigm. This representation, based on the trad-
ition of frames and semantic nets, provides such facilities as
automatic reclassification, value caching, and multiple knowledge
bases. The decisions which led to the choice of this representa-
tion scheme, and the influence that the representation had upon
the development of the FAME system for Financial Marketing Exper-
tise will be discussed.
--
Jim Clarke -- Dept. of Computer Science, Univ. of Toronto, Canada M5S 1A4
(416) 978-4058
BITNET,CSNET: clarke@csri.toronto.edu CDNNET: clarke@csri.toronto.cdn
UUCP: {allegra,cornell,decvax,linus,utzoo}!utcsri!clarke