Download presentation
Presentation is loading. Please wait.
1
Research Information Session Associate Professor John Thornton Gold Coast BIT Honours Degree Convenor and Research Higher Degree Convenor J.Thornton@griffith.edu.au
2
John Thornton Research with Dr John q Gold Coast Honours Convenor q Associate Director IIIS for Gold Coast q RHD Coordinator IIIS and ICT Gold Coast q NICTA researcher q Leader of Constraint Satisfaction and Hierarchical Temporal Memory research groups 8 PhD completions 1 MPhil, 2 Masters, 5 Honours completions
3
John Thornton What are Constraints? A constraint is a relationship over object(s) in the world. What is allowed? What is not allowed? Knowledge about broad range of real world domains can be easily expressed in terms of constraints
4
John Thornton Constraint Programming “Constraint programming represents one of the closest approaches computer science has yet made to the Holy Grail of programming: the user states the problem, the computer solves it.” Eugene Freuder.
5
John Thornton Constraint Satisfaction Given: q A set of variables q A set of permitted values for each variable q A set of constraints on subsets of variables Find: an assignment of values to variables such that all the constraints are satisfied.
6
John Thornton General Techniques q Problems are often NP-complete q Over-constrained q Two classes of technique: Backtracking Local search
7
John Thornton Eight Queen Problem Variable Constraint Domain
8
John Thornton Local Search Place 8 Queens randomly on the board Pick a Queen: Calculate cost of each move 31054111 Take least cost move then try another Queen 04441111 434111131 333211112 344111132 223422212 323211231 204212231 232213231 233212221 232322131 223213211 322333301 Answer Found
9
John Thornton Selected Results q Building Structure into Local Search for SAT IJCAI’07 Distinguished Paper Award q Winner of SAT Competition Gold Medals gNovelty+ (2007 & 2009), R+AdaptNovelty+ (2005) q Temporal Reasoning Local Search (JLC), New SAT encoding (CP’06 & AIJ 2008) q Hybrid Search Resolution + SLS (AAAI’05) q Evolving Algorithms for CSPs Genetic programming (CEC’04, PRICAI’04)
10
John Thornton Practical Problems q NICTA ATOMIC Project Building the local search component of the G12 constraint solving platform q NICTA Vehicle Routing Project Logistic planning of vehicles, crews, deliveries and pick-ups q Jeppesen (Boeing) Local search for postal delivery problem q SAT Competiton Parameter free search, portfolio algorithms, exploiting structure q Glycomics and Eskitis Protein folding for malaria research Identifying pathways in Parkinson’s disease
11
John Thornton Machine Learning q Hierarchical Temporal Memory Using insights from computational neuroscience to build more robust and flexible pattern recognition machines Exploiting temporal connections between inputs (temporal pooling) Combining recognition with prediction
12
John Thornton The Teams CSP/SAT: NICTA QRL: Abdul Sattar, Duc Nghia Pham, Hakim Newton Jeppesen: Valnir Ferreira Jr. NICTA Sydney/Canberra: Michael Maher, Andrew Verden, Phil Kilby IIIS HTM: Michael Blumenstein, Trevor Hine, Jolon Faichney
13
John Thornton Thank You Questions? (also see www.cit.gu.edu/~johnt/)
Similar presentations
© 2025 SlidePlayer.com. Inc.
All rights reserved.