|
Research projects are available for students at Third Year, Honours and
Ph.D. levels. I encourage students to participate in all aspects
of research, including identifying a project. The goal is to locate
a question that you
find interesting. The question may be completely novel or part of a
program of ongoing research, provided I think that we can answer it!
Research in this area will develop skills (computation, numerical analysis,
modelling, visualisation, data analysis, statistical techniques) which
are highly transferrable and attractive to employers.
Candidates will have the
opportunity to collaborate with experts at world class institutions with
whom I maintain ongoing research partnerships,
including Lockheed Martin and Stanford University.
My research is published in high-impact peer reviewed journals
such as the Astrophysical Journal, the premier US journal in the field.
You can achieve research at this level!
Other projects: Current projects will be listed below.
A variety of other projects in related areas are possible.
PhD project: Generic methods in data mining
Mike Wheatland, SIfA/The University of Sydney
(m.wheatland@physics.usyd.edu.au)
Stephen Hardy, NICTA (Stephen.Hardy@nicta.com.au)
Natural scientists are increasingly confronted with problems in the
analysis of large, multi-parameter datasets. For example, astrophysical
images of objects may be available for multiple times at different
wavelengths, or geophysical data sets may be available at different
spatial locations. This project will develop new methods applicable to
combining data from different data sources to improve the accuracy of
the information inferred from the data. These methods will be suitable
for use across the natural sciences, as well as related areas such as
finance and big data analytics, based on the latest machine learning
algorithms (e.g. this paper).
The methods will be tested
on specific problems, e.g. multi-wavelength solar observations, and the
inversion of large scale geophysical datasets such as gravity and
magnetic data. The project requires a strong maths and statistics
background, programming skills, and interest in inter-disciplinary
applications. The successful applicant's supervision will be provided
jointly at NICTA and at SIfA/The University of Sydney and NICTA will
contribute a top-up to a student's PhD scholarship (see
this page for details).
PhD project: Model identification for nonlinear systems via machine
learning
Mike Wheatland, SIfA/The University of Sydney
(m.wheatland@physics.usyd.edu.au)
Stephen Hardy, NICTA (Stephen.Hardy@nicta.com.au)
Is it possible to determine a model, i.e. a set of governing dynamical
equations, based only on observing the time-behaviour of a nonlinear -
and possibly chaotic - system? Recent work (see
this page)
demonstrates that
this is possible for specific problems, using machine learning methods.
In this project new techniques will be developed for tackling this
problem, using the latest developments in non-parametric Bayesian
analysis and machine learning. Specific applications will be
investigated - e.g. modelling the underlying dynamics of non-linear time
series data from chaotic lasers, modelling the dynamical behaviour of
sunspots, and modelling the evolution of forest ecosystems using
ecology datasets. The project requires a strong maths and statistics
background, as well as programming skills. The successful applicant's
supervision will be provided jointly at NICTA and at SIfA/The University
of Sydney and NICTA will contribute a top-up to a student's PhD
scholarship (see this page
for details).
Hons project: Solar flare prediction via non-parametric Bayesian
estimation
Mike Wheatland, SIfA/The University of Sydney
(m.wheatland@physics.usyd.edu.au)
Stephen Hardy, NICTA (Stephen.Hardy@nicta.com.au)
Solar flares are explosive events involving magnetic energy release in
the Sun's corona. Because flares directly affect the Earth's space
weather environment (leading e.g. to damage to satellite electronics),
there is considerable interest accurate solar flare prediction. However,
the flare mechanism is not well understood, and existing methods of
prediction are probabilistic, generally relying on imperfect
associations between properties of sunspot regions in which flares occur
(in particular properties of the observed magnetic field of the sunspot
regions producing flares), and flare occurrence. A Bayesian approach to
flare prediction based only on observed solar flare statistics has been
developed and tested (Wheatland 2004, 2005). In this project the event
statistics method will be combined with machine learning approaches to
forecasting, allowing incorporaton of the additional information
provided by the properties of flare producing regions on the Sun. There
is scope for analytic and numerical work. The project will be jointly
supervised by Stephen Hardy at NICTA, who has expertise in machine
learning.
References:
Wheatland, M.S. 2004, Astrophysical Journal 609, 1134-1139
Wheatland, M.S. 2005, Space Weather Vol. 3, No. 7, S07003
doi:10.1029/2004SW000131
Hons/UG project: Bayesian analysis of large solar flare occurrence
Mike Wheatland, SIfA/The University of Sydney
(m.wheatland@physics.usyd.edu.au)
In 1859 Richard Carrington observed a gigantic solar flare on the Sun,
which subsequently disabled telegraph communications on the Earth for
about a day. Solar flares are
magnetic explosions in the solar atmosphere that affect our local
"space weather." The 1859 Carrington flare disrupted
the communications technology of the day, and in the modern era
we face the prospect of the loss of communications satellites due
to a huge space weather storm (estimates of the cost run to tens
of billions of dollars). It is crucial to know how likely is the
occurrence of a Carrington-scale event. This project will begin
with an investigation of
the rate of occurrence of the largest flares and the time variation
of the rate, using a new Bayesian rate-determination algorithm applied
to data. The project emphasizes physical reasoning and estimation,
programming, data analysis, numerical methods, and statistical
procedures including Bayesian methods.
References:
Wheatland, M.S. 2004, Astrophysical Journal 609, 1134-1139
Wheatland, M.S. 2005, Space Weather Vol. 3, No. 7, S07003
doi:10.1029/2004SW000131
Additional Honours projects:
Additional projects for 2014
are listed in this
document.
Feel free to discuss these and other research opportunities
with me, in person or via e-mail.
Return to Mike Wheatland's home
page
|
