[ont.events] MASTER'S THESIS PRESENTATION

wlrush@water.waterloo.edu (Wenchantress Wench Wendall) (04/28/89)

DEPARTMENT OF COMPUTER SCIENCE
UNIVERSITY OF WATERLOO
SEMINAR ACTIVITIES

MASTER'S ESSAY PRESENTATION

                    -Tuesday, May 2, 1989

Mr.  Guno  Sutiono, graduate student, Dept. of Computer
Science,  University  of  Waterloo,  will  speak  on "A
Region-Based Approach for Digital Cel Painting."

TIME:                 3:30 PM

ROOM:              DC 1304

ABSTRACT

                    Recently,  a  number  of  computer-
aided  paint systems have been developed to improve the
colour  painting process, traditionally one of the most
time-consuming  steps  in 2-D cel animation production.
However,  much  improvement  remains  to be made in the
design of these new systems.

                    In   this   talk,   we  describe  a
region-based   cel   painting  design  which  uses  the
"paint-by-number"  method  and  is  based  on a virtual
frame  buffer  model.  In this region-based design, the
regions are recognizable, defined entities. This allows
us   to   define   functions   for  conceptually-based,
interactive  painting  techniques,  which can be easily
implemented.   Furthermore,  the  availability  of this
additional    region    information    introduces   the
possibility  of  automatic multi-frame colour tracking,
which  can  be implemented by transforming this problem
into  a  well-known  maximum  matching problem in graph
theory.  Thus, this new region-based design builds upon
techniques  used  in  earlier systems and paves the way
for future improvements.

wlrush@water.waterloo.edu (Wenchantress Wench Wendall) (06/15/89)

DEPARTMENT OF COMPUTER SCIENCE
UNIVERSITY OF WATERLOO
SEMINAR ACTIVITIES

MASTER'S THESIS PRESENTATION

                    -Monday, June 19, 1989

Denise  M.  Woit,  graduate  student, Dept. of Computer
Science,   University   of   Waterloo,  will  speak  on
"Rational  Testing for Computational Geometry Codes:  A
Case Study

TIME:                 2:30 p.m.

ROOM:              DC 3540

ABSTRACT

As  the role of software grows in our society, the need
for  reliable and practical software testing techniques
increases.   Functional  testing is one method that may
help  fulfill this need.  While many functional testing
techniques  have  been  theorized,  they  are often too
involved or too complicated to be of practical use.

As  well  as the general testing technique, the testing
process  must  also  be  of  practical use: the testing
environment should be engineered so that the tests will
execute  with  minimal  human interaction and be easily
repeatable.

In  the  area of computational geometry, algorithms are
sometimes  found to be intricate and counter-intuitive;
these  algorithms,  especially, are in need of software
testing  techniques  which are reliable, yet practical.
One  such  algorithm  is that of the visibility polygon
problem.

The  fundamentals  of  a  visibility polygon algorithm,
VISPOL,  are  presented; a functional testing technique
and  environment  are proposed which, while specific to
VISPOL,  can be extended into a general methodology for
the testing of computational geometry codes; an account
of   the   design  and  implementation  of  the  VISPOL
algorithm,    using    current   software   engineering
practices,  is  given; and, the results of the proposed
testing technique and environment applied to the VISPOL
implementation are presented and analyzed.

wlrush@water.waterloo.edu (Wenchantress Wench Wendall) (07/28/89)

Dept. of Computer Science, will speak on 
``Performance Studies in Optical Disk Storage.''

DEPARTMENT OF COMPUTER SCIENCE
UNIVERSITY OF WATERLOO
SEMINAR ACTIVITIES

MASTER'S THESIS PRESENTATION

                    -Thursday, August 3, 1989

Mr.   K.  Hatzilemonias,  graduate  student,  Dept.  of
Computer  Science,  will speak on ``Performance Studies
in Optical Disk Storage.''

TIME:                 1:00 p.m.

ROOM:              DC 1302

ABSTRACT

Optical  disks,  due to the low storage cost per bit of
stored  information,  easy  replication,  large storage
capacity,  and long archival life,  are projected to be
the   dominant  storage  technology  for  many  storage
intensive applications.

In  the  research  presented we investigate the various
types  of  available  optical  disk storage systems and
through  experimentation  we identify their performance
characteristics.

Based  on the experimental results, mathematical models
capturing  the  behavior  of  these  systems  have been
developed and validated against the actual systems they
represent.   These   models  are  incorporated  in  the
implementation  of  simulators for various optical disk
storage systems.

We  investigate the effects of a new notion of blocking
and  we  show that in certain application environments,
it  can  improve  the  retrieval performance of optical
disk storage systems.

Optical   disk  storage  media  is  new  and  not  well
understood. The research presented provides answers and
new  insights  about  the  various optical disk storage
architectures and models their access performance.

wlrush@water.waterloo.edu (Wenchantress Wench Wendall) (07/28/89)

Dept. of Computer Science, will speak on 
``Software Portability by Virtual Machine Emulation.''

DEPARTMENT OF COMPUTER SCIENCE
UNIVERSITY OF WATERLOO
SEMINAR ACTIVITIES

MASTER'S THESIS PRESENTATION

                    -Wednesday, August 2, 1989

Mr.  Stefan  Vorhoetter,  graduate  student,  Dept.  of
Computer  Science, will speak on ``Software Portability
by Virtual Machine Emulation.''

TIME:                 3:30 p.m.

ROOM:              DC 1304

ABSTRACT

The proliferation of diverse computer architectures has
resulted  in  an  increased need for portable software,
but  many  portability  techniques  incur a performance
penalty.   This  penalty  can be reduced by restricting
the  range  of architectures to which a program must be
portable,  at  the  risk of limiting portability to new
hardware as it becomes available.

In  this  thesis  we  examine  a technique for making a
program  portable to architectures for which it was not
intended,  without  affecting  its performance on other
architectures.   The  technique used is virtual machine
emulation,   whereby   an   ideal   (for  the  program)
architecture  is  emulated  by  another  program  on an
incompatible  system.  By designing the virtual machine
carefully,  performance  loss  due  to emulation can be
minimized,   and  the  program  to  be  ported  can  be
recompiled    for    the    virtual   machine   without
modification.

wlrush@water.waterloo.edu (Wenchantress Wench Wendall) (08/11/89)

Dept. of Computer Science, will speak on 
``Tuples in Imperative Programming Languages.''

DEPARTMENT OF COMPUTER SCIENCE
UNIVERSITY OF WATERLOO
SEMINAR ACTIVITIES

MASTER'S THESIS PRESENTATION

                    -Friday, August 18, 1989

Mr.  David  Till,  graduate  student, Dept. of Computer
Science   will   speak   on   ``Tuples   in  Imperative
Programming Languages.''

TIME:              2:00 p.m.

ROOM:              DC 3540

ABSTRACT

Ordered  lists  of  heterogeneous  elements,  known  as
tuples,   appear   in   most   imperative   programming
languages,  usually  as  parameter  lists  for  routine
invocations  or  as  subscript  lists for array element
extraction.   In  this  talk,  other tuple contexts are
developed,  and  three  constructs relating to them are
discussed:  the multiple assignment statement, multiple
return  values  from  routines,  and  input  parameters
passed by means of the assignment statement.

The  programming  language K-W C, which implements most
of  these  tuple contexts and constructs, is introduced
and described.

daemon@watdragon.waterloo.edu (Owner of Many System Processes) (09/06/89)

Dept. of Computer Science, will speak on ``Study of Monitors.''
From: wlrush@poppy.waterloo.edu (Wenchantress Wench Wendall)
Path: poppy!wlrush


DEPARTMENT OF COMPUTER SCIENCE
UNIVERSITY OF WATERLOO
SEMINAR ACTIVITIES

MASTER'S THESIS PRESENTATION

                    -Friday, September 8, 1989

Mr. Michel Fortier, graduate student, Dept. of Computer
Science, will speak on ``Study of Monitors.''

TIME:                 2:00 p.m.

ROOM:                 DC 1304

In  the  design  of  a  concurrent  system, problems of
synchronization  and  communication of processes can be
solve by using monitors.

Over   the   past  15  years several different kinds of
monitors   have  been   investigated.   A  taxonomy  is
presented   that  encompasses  all  these   traditional
monitors   and   suggests   others  not  found  in  the
literature  or  in existing programming languages.  The
classification  scheme  is  based on a single criterion
-  the  scheduling  of  processes   waiting  to use the
monitor which occurs after a wait, after  a  signal and
at  the  exit of the monitor.  It identifies a total of
nine  different  primitive  monitor  types:  six signal
monitors      based      on      categorizations     of
blocking/nonblocking    signal    and    priority    of
resumption,  two  based  on immediate return signal and
priority  of resumption, and two based on an automatic-
signal mechanism and priority of resumption.

An  attempt  is  made to give some empirical discussion
and   analytical   results   about  the  classification
criterion   that  suggest  that  some   monitor  types,
although   functionally  equivalent,  are  better  than
other  types.   In particular, the priority-nonblocking
monitor  has  properties that suggest it is better than
the others.

All   the   monitor   types  have been implemented in C
language.  By having the different monitor types in one
language, it was possible to run tests to compare them.

An    animation   facility   has  been  implemented  to
visualize   the  different   types   of   monitors   by
showing   process   movement   on   entry  queues,   on
condition   variables   and  in and out of the monitor.
With  this  tool, it is possible to see the differences
in the monnitors as well as deadlock and bottlenecks.

daemon@watdragon.waterloo.edu (Owner of Many System Processes) (09/07/89)

Dept. of Computer Science, will speak on 
``A Model for Approximate Text Matching.''
From: wlrush@poppy.waterloo.edu (Wenchantress Wench Wendall)
Path: poppy!wlrush

DEPARTMENT OF COMPUTER SCIENCE
UNIVERSITY OF WATERLOO
SEMINAR ACTIVITIES

MASTER'S THESIS PRESENTATION

                    -Monday, September 11, 1989

Mr.   George   Townsend,  graduate  student,  Dept.  of
Computer   Science,   will   speak  on  ``A  Model  For
Approximate Text Matching.''

TIME:                 3:30 p.m.

ROOM:                 DC 1302

ABSTRACT

     This    thesis   introduces    the    concept   of
Approximate  Text  Matching,   a  natural extension  of
approximate string matching.  A  hierarchical model  is
presented   upon which solutions to this problem may be
based.
     Traditional   approaches   to  approximate  string
matching   deal  with   transformations   which,   when
applied   at   the string layer,  cause two strings  to
become   equivalent  at  the   character layer.  If the
transformations   required   to   achieve    this   are
"reasonable"  as  determined  by  some heuristics,  the
two  original strings  are claimed  to be approximately
equal.
     This  traditional  model consists  of two  layers,
the string  layer and  the character  layer, where  the
string  layer  is  viewed as a  sequence of characters.
The   typical   transformations     applied    to   the
string    layer    are  transposition   of  characters,
removal   of  superfluous characters,  and  addition of
missing  characters.   In  the  model   presented here,
these  same notions  are applied at multiple layers  of
the  problem.   "Text"  is   viewed  as  a sequence  of
"fields  of data",  or  phrases. A  "field of data"  is
viewed  as a sequence of "unit strings" or words.  Each
"unit string" is  viewed, in the traditional manner, as
a     sequence    of   characters.   The    traditional
transformations    considered   in  approximate  string
matching  are  applied  at  each  of  these layers in a
general way.
     The  advantages of  the model are twofold.  First,
the  model  understands   the structure  of  the  data,
and    will   therefore   search   for  matches  for  a
particular field in an appropriate place. Secondly, the
failure    of   the  model   at  one  layer   is  often
compensated for at  the next  highest layer.
     The  Oxford  English  Dictionary  was  used  as  a
testing  ground  for  algorithms based upon this model.
Citations  occurring  in  the body were matched against
their  sources  in  the  O.E.D.  bibliography.   On two
samples  of data, correct matches were made over 90% of
the time.

wlrush@water.waterloo.edu (Wenchantress Wench Wendall) (09/22/89)

Dept. of Computer Science, will speak on 
``The Automatic Inbetweening of Keyframes Composed of 
Spline Curves Satisfying Various Constraints.''


DEPARTMENT OF COMPUTER SCIENCE
UNIVERSITY OF WATERLOO
SEMINAR ACTIVITIES

MASTER'S THESIS PRESENTATION

                    -Thursday, September 28, 1989

Mr. Ronald Hardock, graduate student, Dept. of Computer
Science, will speak on ``The Automatic Inbetweening  of
Keyframes  Composed of Spline Curves Satisfying Various
Constraints.''

TIME:                 2:30 p.m.

ROOM:                 DC 1304

ABSTRACT

    This thesis presents a method one can use to obey a
certain class of constraints between spline curves in a
keyframe  inbetweening environment. The general problem
is  ensuring  that  constraints,  which  are  true  for
keyframes, will also be true for the computer-generated
inbetween frames.

    Various  types  of constraints are outlined.  A set
of  conditions on the interpolation method are outlined
such  that  linear  mapping  constraints can be obeyed.
Mention  is  also  made of various constraints that are
not   expressible  as  linear  mappings,  to  point  to
directions of future research.

wlrush@water.waterloo.edu (Wenchantress Wench Wendall) (09/28/89)

Dept. of Computer Science, will speak on 
``Hypermedia Document Information Based Retrieval Applications.''



DEPARTMENT OF COMPUTER SCIENCE
UNIVERSITY OF WATERLOO
SEMINAR ACTIVITIES

MASTER'S THESIS PRESENTATION

                    -Thursday, October 5, 1989

Suzanne  M. Potvin, graduate student, Dept. of Computer
Science,    will   speak   on   ``Hypermedia   Document
Information Based Retrieval Application.''

TIME:                 10:30 p.m.

ROOM:              DC 3540

ABSTRACT

Hypermedia   systems   emphasize   text  and  graphical
information  and  provide  interactive  interfaces  for
browsing  and  editing  information  concepts. However,
most  hypermedia  systems do not provide adequate query
mechanisms.

In  this presentation, we present a specification and a
prototype implementation of a hypermedia document-based
information   retrieval   application   (HyDIRA).  This
application  combines  the advantages of both hypertext
and   document   management   systems,   and  minimizes
disadvantages  related  to  present  hypertext systems.
The   prototype   version  that  has  been  implemented
demonstrates   the   feasibility   of   the  goals  and
functionality provided by HyDIRA.

The  document model used resembles the ODA/ODIF object-
oriented  standard  for  Office  Document Architectures
(although  it  is limited).  We will present some first
feedback on how the ODA/ODIF standard could be extended
to  support  hypertext/hypermedia  applications  in  an
integrated and uniform manner.

wlrush@water.waterloo.edu (Wenchantress Wench Wendall) (10/12/89)

will speak on ``An Evaluation of Tension within an Extensible 
Spline Testing Facility.''



DEPARTMENT OF COMPUTER SCIENCE
UNIVERSITY OF WATERLOO
SEMINAR ACTIVITIES

MASTER'S THESIS PRESENTATION

                    -Monday, October 16, 1989 NOTE!! DATE CHANGE!!

Mr.  Paul  Ruest,  graduate  student, Dept. of Computer
Science,  will  speak  on  ``An  Evaluation  of Tension
within an Extensible Spline Testing Facility.''

TIME:                 10:30 a.m.

ROOM:                 DC 3540  NOTE!! CHANGE OF ROOM

ABSTRACT

    Beta2  and rational B-splines have received a great
deal  of  attention in recent years.  Both associate an
extra parameter called tension with each control vertex
which  allows the curve to be pulled or pushed locally,
relative  to  the  control  polygon, and independent of
control vertex movement.

    An  obvious  question from the design point of view
is whether users actually find these tension parameters
useful.   Therefore  an  experiment  was  designed  and
performed  to  explore  the  usefulness of tension when
interactively designing curves.

    The  experiment was performed on a reworked version
of  an  extensible  spline testing software system that
has  been  developed  at  the  University  of  Waterloo
Graphics lab.

                    October 5, 1989

wlrush@water.waterloo.edu (Wenchantress Wench Wendall) (11/17/89)

DEPARTMENT OF COMPUTER SCIENCE
UNIVERSITY OF WATERLOO
SEMINAR ACTIVITIES

MASTER'S THESIS PRESENTATION

                    -Wednesday, November 22, 1989

Martin  Van Bommel, graduate student, Dept. of Computer
Science,  will  speak  on ``Functional Dependencies and
Join Conditions in Object-Oriented Data Models''

TIME:                 2:30 p.m.

ROOM:                 DC 1331

ABSTRACT

Several   kinds   of   explicit   constraints  for  the
relational model have proven invaluable in appraising a
choice  of  relation schemes, in designing the physical
database,  and  in  optimizing  queries.  Variations of
these  constraints for object-oriented data models will
almost certainly have the same benefits.  In this talk,
I  will discuss two possible forms of constraints for a
simple  choice  of  object-oriented  data  model.  Some
preliminary  results  on  applications of an ability to
reason about these constraints will be examined.  Sound
and   complete   axiomatizations   will  be  presented,
followed  by  an  outline  of  decision  procedures for
several  special  cases  in  which  constraints satisfy
further syntactic restrictions.

                   November 15, 1989

wlrush@water.waterloo.edu (Wenchantress Wench Wendall) (12/01/89)

DEPARTMENT OF COMPUTER SCIENCE
UNIVERSITY OF WATERLOO
SEMINAR ACTIVITIES

MASTER'S THESIS PRESENTATION

                    -Tuesday, December 5, 1989

Mr.   Dexter   Bradshaw,  graduate  student,  Dept.  of
Computer    Science,    will   speak   on   ``Retrieval
Optimization in the Minos Distributed Testbed.''

TIME:                 10:30 a.m.

ROOM:                 DC 1331

ABSTRACT

MINOS  is  a  distributed  data  management system that
supports   the  management  of  complex  data  objects.
Particular  emphasis  is  given  to  the  management of
complex  multimedia  data  (text,  voice, images, video
etc.).     This    thesis    describes    the   design,
implementation and performance evaluation of algorithms
for improving the retrieval performance of the system.

Three types of performance optimizations are described.
Query evaluation performance is improved by introducing
buffering  and  scheduling  algorithms that exploit the
sequential  characteristics  of the access methods used
by  the system for queries involving text and attribute
searches.   Data  retrieval from the database server is
improved  by  reducing  the  disk  exchanges in jukebox
server  architectures  through  appropriate  long  term
scheduling  algorithms.   Finally,  network  traffic is
reduced  by  using  set  processing,  routing  and flow
control  protocols  between  the  data  server  and its
clients.

                   November 30, 1989

wlrush@water.waterloo.edu (Wenchantress Wench Wendall) (12/08/89)

DEPARTMENT OF COMPUTER SCIENCE
UNIVERSITY OF WATERLOO
SEMINAR ACTIVITIES

MASTER'S THESIS PRESENTATION

                    -Thursday, December 14, 1989

Ms.   Yvonne   Dubreuil,  graduate  student,  Dept.  of
Computer  Science,  will  speak on ``The Integration of
Buffer Management and Recovery Management.''

TIME:                 2:30 p.m.

ROOM:                 DC 1304

ABSTRACT

The basics of IBM's ARIES Database Recovery System will
be  discussed.  This system uses fuzzy checkpoints thus
all  the  buffers containing dirty pages do not have to
be  written  out  during  checkpointing.   The  cost of
checkpointing  is significantly reduced especially with
the large buffer pools in use today.

Buffer  Managers  and  Recovery  Managers tend to be at
odds with each other.  The buffer manager wants to keep
pages  it  expects to be accessed in the near future in
the  buffer  pool  while  the recovery manager wants to
write  out  the  pages  that  will  poorly  affect  its
recovery performance.  I will discuss a buffer manager,
ODPF (Oldest Dirty Page First), that bases its criteria
for  page  replacement  on  data  that  is  used by the
recovery  system thus having the effect of lowering the
I/O cost of recovery even further.  Performance results
based on simulation will also be presented.

                    December 7, 1989