Computer Sciences Lab

1
Computer Sciences Lab NICTA Opportunities for
Honours projects 2009Marcus Hutter (adapted
from and thanks to Sylvie Thiebaux)
2
College of Engineering Computer Science (
Friends)
DCS
CSL
NICTA
DEng
InfoEng
3
Research Groups

• Research focuses on
• Artificial Intelligence
• Machine Learning
• Logic Automated Reasoning
• Computer Vision
• 40 researchers in 4 groups
• Diagnosis, Planning, Optimisation (DPO, aka
  KRR)
• Statistical Machine Learning (SML)
• Logic Computation (LC)
• Vision Science, Technology Applications (VISTA)

4
Diagnosis, Planning Optimisation Group

• Diagnosis
• Explain abnormal situations from observations
• Circuits, power networks, web services, humans
• Planning
• Decide schedule the tasks to be undertaken to
  meet given objectives
• Project planning, military operations planning,
  robot control, solving puzzles games
• Optimisation
• Find the best possible solution to a problem
• How can we predict the hardness of optimisation
  problems?

Contact Jussi.Rintanen_at_nicta.com.au
5
Model-Based Supervision of Composite Systems
Composite systems feature simple components
organised into a highly reconfigurable
architecture
Examples web grid services, power and water
systems telecom networks, traffic control systems

• Supervision tools confer the ability to
• self-diagnose to detect faults in the system and
  explain their causes
• self-reconfigure to restore or improve service

Project goals develop theories, algorithms
tools for the supervision of composite systems
Approach draws on artificial intelligence (model-
based diagnosis, planning), discrete- event
systems, and model-checking
Contact Jussi.Rintanen_at_nicta.com.au
6
Statistical Machine Learning Group

• Machine Learning automates the input-output
  mapping.
• Lots of fun projects for analysing data.Let us
  do both theory and application

input(data)DocumentsVideoMoleculesMicroarrays
Sensor NetworksMission Plans
output(analysis) Authors, scriptPeople,
scenesBiological function Cancer
diagnosisNovelty, alarmOptimal strategy
Magic happens
Contact Wray.Buntine_at_nicta.com.au
7
Document Analysis
Suffix tree

• .
• Build document similarity measure
• Build fast discriminative optimiser (SVM style)
• Integrate into mail filtering system (e.g. DSPAM)

c
Spam filter
ab
b
c
abc
c
abc
Contact Wray.Buntine_at_nicta.com.au
8
Logic and Computation Group

• Logical analysis of systems
• Assure correctness, safety, robustness
• Software systems (are votes counted okay?)
• Physical systems (will the robot arm break?)
• Trust and Security (can I trust this eBay
  seller?)
• Tools for reasoning by computers
• Logical deduction Does it follow?
• Constraint satisfaction How might it be?
• Theory behind all this
• New kinds of logic for new tasks

Contact Rajeev.Gore_at_rsise.anu.edu.au
9
Constraint Satisfaction Platform (G12)

• Constraint Satisfaction Problem
• Hard constraints - e.g. every team plays every
  other at home and away
• Soft constraints - e.g. fairness conditions
  (may be complex)
• Additional requirements from TV stations, etc.
  complicate further
• Difficult computational problem

Contact John.Slaney_at_anu.edu.au
10
L4 Verified
L4 Micro-kernel L4 operating system used in
embedded systems e.g. sensor networks, mobile
phones Small trusted kernel (guarantees
separation properties) NICTA project formally
verify the kernel Project runs until 2008 One of
the most ambitious formal verification projects
ever undertaken anywhere Commercial potential if
successful
Contact Michael.Norrish_at_nicta.com.au
11
Vision Science, Tech. Applications Group

• Major projects
• Spectral imaging
• Smart cars
• Medical image analysis
• Surveillance

Contact Nick.Barnes_at_nicta.com.au
12
Smart Cars
Contact Lars.Petersson_at_anu.edu.au
Pedestrian detection tracking
Speed sign detection recognition
Car detection tracking
A complete driver assistance system, focusing on
driver safety
13
Automatic Anatomical Structure Extraction

• Detection of Alzheimers disease
• changes to hyppocampus implicated- doctors
  hand-trace each scan slice- obtain a math.
  representation for analysis- need to repair and
  parametrise the 3D data

Contact Paulette.Lieby_at_nicta.com.au
Parametrisation
Topology repair
14
Artificial Intelligence

• Universal Artificial Intelligence
• 
  
• Decision Theory Probability Utility Theory
• 
  
• Universal Induction Ockham Bayes Turing
• Information-theoretic,
• Statistical, and
• Philosophical,
• Foundations of
• Artificial Intelligence

Contact Marcus.Hutter_at_anu.edu.au
15
Finally

• These slides are at http//www.hutter1.net/rsise/
  honours.ppt
• Many other projects, for exmple in
• Traffic control Philip.Kilby_at_anu.edu.au
• Game playing Adi.Botea_at_nicta.com.au
• Agent architectures John.Lloyd_at_anu.edu.au
• Artificial AI, Trust Jochen.Renz_at_anu.edu.au
• Automated deduction Rajeev.Gore_at_anu.edu.au,
  John.Slaney_at_anu.edu.au, Peter.Baumgartner_at_nicta.co
  m.au
• Satisfiability Anbulagan_at_nicta.com.au,
  Jinbo.Huang_at_nicta.com.au, Jussi.Rintanen_at_nicta.com
  .au
• Security protocol verification Alwen.Tiu_at_anu.edu.a
  u
• If you like theory Marcus.Hutter_at_anu.edu.au
• Apply for a summer scholarship with us !