Centre for Research in Intelligent Systems

Vacation Scholarships 2009

The Centre is offering three vacation scholarships.  These will be awarded on a competitive basis.  Details of the scholarships can be found here and here.  The application closing date is 5:00pm 9 October 2009.  The application form can be found here.

Information on the centre can be found here.

The scholars will be able to chose between the following projects. 

Interactive layout of family trees and other hierarchical diagrams using Dunnart.

Supervisors: Kim Marriott & Michael Wybrow

Dunnart is an open source constraint-based  interactive diagramming editor developed by Michael Wybrow. It provides automatic layout that the author can customize by moving objects around and adding placement constraints. The project is to extend Dunnart to support hierarchical diagrams like family trees.

Learning Dynamic Bayesian Networks

Kevin Korb & Ann Nicholson

Dates: December - Jan

Pre-req: FIT3080 or equivalent background

Static Bayesian networks (BNs) represent causal processes which may be
repeated, but which do not change upon repetition. While simple in
concept, they are nevertheless powerful representations for dealing
with systems showing significant uncertainty about their outcomes, and
they are widely employed in health care, ecological and environmental
sciences, manufacturing industries, and government (see
www.norsys.com/clients.htm). A great deal of effort has been put into
automating the learning of Bayesian networks from data, including at
Monash, yielding our CaMML program (Causal discovery via MML).

Dynamic Bayesian networks (DBNs) represent causal systems which
change over time, yielding time series data from their
observation. Examples include all kinds of economic markets, such
as stock markets and currencies, weather and climate systems, and human
disease processes. DBNs extend the range of application of BNs in important
ways; this project will implement and test simple mechanisms for
extending CaMML to learn them from time series data.

A software framework for generalised cylinder construction

Supervisor: Jon McCormack

Generalised cylinders are a geometric representation method originally developed for robotic vision and planning. For this project we are interested in developing a framework for the specification of biological form using generalised cylinders. This method is well suited to a generative, procedural specification and will be linked to a developmental model for generating biological form. The software developed will form a C++ framework developed for interactive media applications. The successful student for this scholarship should be familiar with basic concepts in computer graphics and geometric modelling and also have experience in C++ programming.

Classifier Systems for the CEMA Agent Modelling Framework

Supervisor: Jon McCormack

The CEMA Agent Modelling Framework is a C++ API for agent-based modelling experiments in Artificial Life. The API consists of a series of libraries that cover basic Agent modelling, GUI, Audio and Scheduling tasks. This project will extend the framework by adding Classifier Systems as a mechanism for agent learning. Classifier Systems provide agents with a mechanism to learn and adapt their behaviour in response to their environment.  The project will implement a variety of algorithms including XCS, ZCS and XCSMH, and M-LCS.

For a overview see:


Text clustering based concept hierarchy to generalize from different text sources

Supervisors: Damminda Alahakoon

Humans accept inputs from many different sources and learn incrementally from those sources.  The accumulated information is used in performing many activities. In the same way, if we look into the field of text mining, there is variety of sources of information representing the related things. Therefore building a central information processing unit that merges all the different sources would be more advantages in information processing and making decisions. 

Technically, we can cluster different text sources and build a central incremental conceptual hierarchy merging the clustered results. In detail, there should be an ongoing central concept hierarchy that is adapted with the inputs from different sources. The clustered results should be merged to the hierarchy so that it forms the generalized view of all the input information sources. There are several text clustering algorithms that could be used for this purpose. Also there are several concept hierarchy building techniques. Information about these techniques will be provided to interested parties.

The main steps involving this process can be listed down as below.

  1. Clustering the different text sources and derives the concepts represented by them.
  2. Merge the concepts into a hierarchy using concept building techniques.

Such a system can have huge potential value in many fields. Information from emails, other electronic documents and web sites identified by web crawlers could be put together to obtain an understanding of related fields, subfields etc as well as what words, concepts are related together.

A simulation investigation of sensors that could be used to assist Museum visitors

Supervisors: David Albrecht, Ingrid Zukerman and Geoff Webb

Imagine when you next visit a museum you are handed a device that not only tells you about the current exhibit that you are viewing but also makes recommendations about other exhibits that it believes you may be interested in. For such a device to be useful it will need to interact with sensors that are keeping track of your movements, and it will need to make predictions about your trajectory and interests. There are are several types of sensors, and it is too expensive to try each of them out in a real museum environment.

This summer project is a continuation of an existing project, where we propose to simulate the effect of different types of sensors on modelling visitors behaviour in a museum.

You must be a competent programmer and familiar with MATLAB (or at least familiarize yourself with MATLAB prior to the start of the project).

DORIS (Dialogue Oriented Roaming Interactive System)

Supervisors: Ingrid Zukerman, Gideon Kowadlo and Patrick Ye

In this project, we apply a probabilistic approach to develop a dialogue module for a robot. The project consists of the following modules: (1) speech recognition, (2) language interpretation, and (3) dialogue. Two post-doctoral fellows are currently working on different aspects of this project.

Possible summer vacation projects are:
* a software component that takes information obtained by a simple vision system and gesture recognition system (provided by Prof R Jarvis from Engineering), and combines it with the output of the language interpretation component.

* a speech recognition server that takes spoken input in real time and interacts with the rest of the language interpretation module (at present speech is being recorded off line). This server could be combined with a web-interface under development which shows
the robot in a virtual environment.

The student would have to be a competent programmer, and have solid mathematical skills.

Interested students should contact Dr Gideon Kowadlo or Dr Patrick Ye


Supervisors: Ingrid Zukerman and Adrian Bickerstaffe

In this project, we apply statistical techniques to determine the overall sentiment and specific ratings in review articles (e.g., film or product reviews). One post-doctoral fellow and one PhD student are currently working on different aspects of this problem.

The summer vacation student would study the effect of collapsing similar words and phrases on the accuracy of the algorithms being investigated, e.g., ``the movie was bad'', ``the film was bad'', ``the film leaves something to be desired''.

The student would have to be a competent programmer, and have solid mathematical skills.

Interested students should contact Dr Adrian Bickerstaffe:


Supervisors: Ingrid Zukerman, David Albrecht, Geoff Webb, Fabian Bohnert and Patrick Ye

This project is linked to the following project, and could be done as a follow-up of this project:
A simulation investigation of sensors that could be used to assist Museum visitors.

In this project, we apply statistical techniques to infer visitors' trajectory and interests in a museum. The aim is to make recommendations for items to be visited, and generate personalized summaries of a person's visit. One post-doctoral fellow
and two PhD students are currently working on different aspects of this project.

The summer vacation student would put together a demo-able museum recommender prototype (hardware + software, possibly partly mock-up), with the objectives of (1) pulling together and integrating all the working pieces of code from the various angles of the KUBADJI project on the back-end, and (2) developing a neat, user-friendly front-end.

Interested students should contact Mr Fabian Bohnert or Dr Patrick Ye

Data mining

Supervisor: Geoff Webb

With the massive growth of data storage, there is strong demand for techniques for extracting information from data.  This project will implement advanced data mining software and apply that software to complex problems.  The exact problems to be tackled will depend on the background and interests of the successful scholar.

For further details please contact Prof Geoff Webb, webb@infotech.monash.edu.au