Student Projects

The following projects appropriate for undergraduate, Masters (MCOMP), PhD students and visiting (summer) scholars. Please email the provided contact if you are interested in a project.

If you are interested in doing a PhD with an advisor in the SML, please email Tiberio Caetano (first.last@nicta.com.au) as well as any SML staff members whose research interests you.

For Summer Scholar Projects, please apply to CECS at ANU by 31 August 2008 to complete a Summer Scholar project.

  • Natural Language Query Answering
    Supervisor: Dr Scott Sanner
    Have you ever wanted to ask your computer a question in natural language and have it reply with the correct answer? Build a small prototype using reinforcement learning for natural language query answering in Wikipedia.
  • Finding Latent Variables in Reinforcement Learning
    Supervisor: Dr Scott Sanner
    Examine automated feature induction methods via latent variable learning, for a task in a simulated environment such as learning to play a game, learning to make inferences for logical query answering, and learning to optimize program control structure.
  • Combining Machine Learning and Logic
    Supervisor: Dr Scott Sanner
    Choose a task and implement a small learning and inference engine based on an appropriate knowledge and conditional random fields (CRF), using Java tools.
  • Mars Rovers and Traffic Controllers
    Supervisor: Dr Scott Sanner
    Choose one of the above program domains, and implement an optimal planning system for this task. This project offers the chance to learn about the theory of optimal sequential decisionmaking and its application to practical problems.
  • Scheduling for a High-performance Computing Cluster
    Supervisor: Dr Scott Sanner
    A partner at the ANU is looking for better process schedulers for a supercomputing system. Perform data analysis on server longs, develop a model of process performance, and then feed that data to a constrained scheduler.
  • Universal Artificial Intelligence
    Supervisor: Dr Marcus Hutter
    The dream of creating artificial devices to reach or outperform human intelligence is an old one; investigate some of the hundreds of open questions of universal artificial intelligence.
  • Human Knowledge Compression Contest
    Supervisor: Dr Marcus Hutter
    Investigate the problem of compressing the first hundred megabytes of Wikipedia better than your predecessors.
  • On the Foundations of Inductive Reasoning
    Supervisor: Dr Marcus Hutter
    Humans learn from experience; how should inductive inference be completed? Consider some open problems such as the black raven paradox, the zero p(oste)rior problem, reparametrization invariance, old-evidence, and updating problems, and study Solmonoff's theory of universal induction.
  • Universal Induction versus No Free Lunch
    Supervisor: Dr Marcus Hutter
    Solmonoff's theory based on Occam's razor essentially specifies a universal model class and prior. But the No Free Lunch theorem states that all optimization or search algorithms are equally good/bad, if a uniform average over the space of all functions is taken. Investigate this ongoing battle of theories.