Please feel welcome and invited to distribute this within your
departments/schools.
- - ------------------- ------------------------ ------- - -
PhD scholarship: Statistics and Computer Science
Rating and ranking sports players and teams using Minimum Message
Length
[Australian Postgraduate Award (Industry): APAI
Australian Research Council (ARC) Linkage Project LP100200865
Chief Investigator: David Dowe, Monash University
Industry Partner: Cadability Pty Ltd]
Location: Clayton School of Information Technology, Monash
University, Clayton (this is a suburb of Melbourne), Australia
Starting date: May 2011 or as soon as possible thereafter
Title: Rating and ranking sports players and teams using Minimum
Message Length
Project Background:
Minimum Message Length (MML) is a unifying principle in machine
learning (computer science, artificial intelligence), statistics,
econometrics, inductive inference and what many nowadays seem to
call ``data mining''.
With its origins in Bayesian information theory (Wallace &
Boulton, Computer J [Oxford Univ Press], 1968), MML is now
regularly published in the statistics, philosophy and
econometrics literature - including (e.g.)
J Royal Stat Soc (B) (1987a, 1987b, 1992),
Brit J Philos Sci (2007, pp709-754),
Handbook of Philos of Science - Vol. 7, Phil of Statistics (2010,
pp901-982), etc.
The information-theoretic underpinnings of MML not only make it a
statistically invariant Bayesian form of model selection and
point estimation, but they also make it readily amenable to
combining the discrete and the continuous, such as (e.g.) the
hybrid Bayesian networks in Comley & Dowe (2003) and
Comley & Dowe (MIT Press, April 2005). [These two papers,
fololowing on Dowe & Wallace (1998), are the first papers
combining both discrete- and continuous-valued attributes in MML
Bayesian networks.]
All of this is alongside practical applications in a broad range
of areas.
The formal relationship between Kolmogorov complexity and MML is
perhaps best described in Wallace & Dowe (1999a, Computer J,
pp270-283) or in Wallace (2005, chap. 2) - although it has
certainly been discussed in other places.
MML is especially effective when there is much noise in the data,
model misspecification and/or (as in the Neyman-Scott (1948) case
or the case of [single and especially multiple] latent factor
analysis - e.g., IQ estimation and octane rating estimation) the
amount of data is comparatively small compared to the number of
parameters to be estimated.
Project:
The project is on ``Rating and ranking sports players and teams
using Minimum Message Length''. Rating systems go back at least
as far as Harkness (1949) and the better-known Elo (1961) system
for rating chess players. More recent attempts have been made to
refine these systems in a variety of ways. We will refine the
systems further - perhaps starting with chess but certainly going
much further. This includes dealing with the challenging
(Neyman-Scott-like) situation where, for some players and teams,
there are few games per player or few games between different
groups of players. Our enhanced modelling will be for a range of
games and sports - including advantages such as, e.g., first move
(as in chess), home ground and location, surface (as in tennis),
etc. We will apply this to rating and ranking individuals and
teams. We also refine how quickly ratings can change depending
upon the strength of the player. All sorts of games and sports
could use such better systems for rating and ranking teams.
Applicant background: Applicants should have a background -
including completing at least the equivalent of an undergraduate
degree - in at least one of mathematics, statistics, computer
science and/or (information theory and) electrical engineering.
The successful candidate will have an undergraduate degree and
will be at least semi-literate in at least one of mathematics
and information theory, or at least interested in both areas.
Some experience in computer programming in at least one programming
language is highly desirable.
Applicants should also be able to write computer programs -
preferably in a variation of (e.g.) C or Java.
If applicants consider themselves not to be strong at mathematics,
then they should at least be fond of mathematics.
Further reading: See C. S. Wallace (2005)
www.csse.monash.edu.au/~dld/CSWallacePublications#MMLBook
by the originator of MML.
For even further reading, see one of (e.g.)
Dowe, Gardner & Oppy (Brit J Philos Sci, 2007, pp709-754)
[rated A*],
or
Dowe (2008a, ``Foreword re C. S. Wallace'', Computer J [Oxford
Univ Press], pp523-560) [as guest editor of the Christopher
Stewart WALLACE (1933-2004) memorial special issue]
or
Dowe (2010a, Handbook of Philosophy of Statistics, pp901-982)
[e.g., at
www.csse.monash.edu.au/~dld/David.Dowe.publications.html ].
[Note of interest: www.chessbase.com/newsdetail.asp?newsid=7114 .]
Salary: Standard PhD scholarship (Aus$26,667p.a.) [possibly
tax-free] accompanied by additional top-up.
The scholarship is for the official Monash University standard
duration of 3 years, although this might possibly be extended for
a further 6 or possibly 12 months.
Starting date: May 2011 or as soon as possible thereafter.
Brief note about the sole Chief Investigator (David L. Dowe) :
In 2005, a book on MML by Chris Wallace (1933-2004), the
originator of MML in Wallace & Boulton (Computer J, 1968), was
published posthumously. In the Wallace (2005) book, Dowe is the
most cited living person in the reference list, Dowe has the most
number of pages devoted to his work of any living person, Dowe is
the most mentioned living person in the table of contents (sec.
4.10 and sec. 4.12.2 both mention his name in their titles) and
Dowe is individually singled out for special gratitude in the
preface on page vi.
- - ----------------------- - -
Supervisor's credentials include (e.g.) :
-----------------------------------------
In Chris Wallace (1933-2004)'s posthumous ``Statistical and Inductive
Inference by Minimum Message Length'' (2005),
(a) I am given special mention in the preface on page vi,
(b) I am the only living person mentioned in the table of contents,
where my name appears twice,
(c) I am the living person whose name and work are most mentioned in
the index,
(d) other than Chris Wallace himself, (in the reference list) I am
the most cited author.
Wallace & Dowe (1999a) was once the Computer J (OUP)'s most
downloaded article - and currently remains as Chris Wallace's most
cited co-authored work by a researcher still active in the area.
Hernandez-Orallo & Dowe (2010) is currently the (A* rated)
Artificial Intelligence J's most downloaded article, recently
featured (5/March/2011, page 82) in The Economist magazine (to name
one of 60+ pieces of media coverage).
I co-authored the first papers on MML Bayesian nets which combine
both discrete (multi-valued) and continuous-valued attributes.
I have proved [Dowe (2008a, 2008b, 2011)] a uniqueness result about
the invarian ce of log(arithm)-loss probabilistic scoring.
[This follows on my publishing papers with probabilistic models and
log-loss scoring since 1993 - well before many people saw the merits
of doing probabilistic classification.]
I have been invited to contribute a piece for the forthcoming
Handbook of Philosophy of Statistics.
Etc.
[Please forgive the not so self-effacing section immediately above,
but recent circumstances seem to render it necessary. (That said,
please let me know if you can't find any of the abovementioned
references or if you'd like further reading or other details.)]
- - ----------------------- - -
For prospective applicants:
[Please understand that I expect many enquiries and applications,
so the more clearly applicants express themselves and the easier
they make my life :-) , the higher their probability of getting
due attention.]
Enquiries: Contact
David dot Dowe arroba infotech dot monash.edu dot au
www.csse.monash.edu.au/~dld/David.Dowe.publications.html
(www.Solomonoff85thMemorial.monash.edu.au)
with clear e-mail subject line and contents.
Applications: Cover letter and application - as above - including
academic transcripts, proof (if appropriate) of English language
quality (e.g., IELTS or TOEFL) [as this is a Monash University
requirement], addressing selection criteria, and including
c.v./resume'.
For information about English language requirements [in
``Information Technology''], see
www.mrgs.monash.edu.au/futurestudents/eligibility/langprof.html .
If applying, please send application once - in one e-mail - and
please include every relevant file as a separate attachment.
Again, send to
David dot Dowe arroba InfoTech dot monash.edu dot au
www.csse.monash.edu.au/~dld/David.Dowe.publications.html
(www.Solomonoff85thMemorial.monash.edu.au)
with clear e-mail subject line and contents.