You are here: School of Geosciences / People / Academic Staff

Dr Thomas Landgrebe

Postdoctoral Fellow, EarthByte Group

Madsen Building (F09), Room 410
Phone: +61 2 9036 7175
Fax: +61 2 9351 2442
Email:

Research Interests

Thomas is currently working in the School of Geosciences as part of the EarthByte group to provide data-mining and analysis research and development at the interface between Geophysics, Maths and Software Engineering. His focus is on analysing data associated with tectonic plates, and reconstructed back through geological time. This will manifest itself via integration of a number of analysis tools into the GPlates framework, aiding knowledge discovery and exploration for Geoscience researchers.

Thomas specialises in statistical pattern recognition, specifically the systematic design of large multiclass classification systems, and its related topics: working with multiclass operating characteristics, combining base classifiers and feature sets, advanced evaluation methods, and finally, how to incorporate domain knowledge into the multiclass system to make better trade-offs. He has worked in industry in the areas of statistical pattern recognition, image processing, hyperspectral imaging, and video processing, with major projects involving automated mineral sorting, 3-dimensional particle analysis, and intelligent CCTV surveillance. Thomas completed his doctorate at Delft University of Technology in the Netherlands in 2007, and has also completed MSc and BSc electrical engineering degrees at the University of the Witwatersrand in South Africa in 2003 and 1999 respectively.

Selected Publications

  • Paclik, P., Lai, C., Landgrebe, T.C.W., and Duin, R.P.W., ROC analysis and cost-sensitive optimization for hierarchical classifiers, 2010 International Conference on Pattern Recognition (ICPR 2010), Turkey 2010
  • Landgrebe, T.C.W., Müller, R.D., and Williams, S., Knowledge discovery via a virtual geological observatory, eResearch Australasia 2010, 8-12 November 2010, Gold Coast, Australia
  • Landgrebe, T.C.W. and Pacik, P., The ROC skeleton for multiclass ROC estimation, Pattern Recognition Letters, January 2010, doi: 10.1016/j.patrec.2009.12.037
  • Landgrebe, T.C.W. and Voigt, A., Multi-camera 3D particle analysis for mineral processing, Anglo American Smart Ideas Conference 2008, August 2008
  • Morrison, G.M., Price, M., Voigt A. and Landgrebe, T.C.W., Incorporating shape in DEM simulation, Computational modelling in Mineral Processing 2008, Cape Town, South Africa
  • Landgrebe, T.C.W. and Duin, R.P.W., Efficient multiclass ROC approximation by decomposition via confusion matrix perturbation analysis, IEEE Transactions on Pattern Analysis and Machine Intelligence, vol 30, number 5, May 2008
  • Landgrebe, T.C.W., Operating characteristics for the design and optimisation of classification systems, PhD Thesis, Delft Technical University, 2007, ISBN 978-90-9022565-4
  • Landgrebe, T.C.W., Mineral classification via hyperspectral imagery using advanced statistical pattern recognition, Anglo American Smart Ideas Conference 2007, August 2007
  • Landgrebe, T.C.W. and Duin, R.P.W., A simplified volume under the ROC hypersurface, SAIEE Africa Research Journal, pages 94-100, vol 98, number 3, 2007
  • Landgrebe, T.C.W. and Duin, R.P.W., Approximating the multiclass ROC by pairwise analysis, Pattern Recognition Letters, pages 1747-1758, vol 28, number 13, 2007
  • Landgrebe, T.C.W. and Duin, R.P.W., A simplified extension of the Area under the ROC to the multiclass domain, Seventeenth Annual Symposium of the Pattern Recognition Association of South Africa, November 2006
  • Landgrebe, T.C.W., Bradley, A.P., Paclik, P. and Duin, R.P.W., Precision-Recall Operating characteristic (P-ROC) curves in imprecise environments, 18th International Conference on Pattern Recognition (ICPR 2006), Hong Kong, China, August 2006
  • Landgrebe, T.C.W. and Duin, R.P.W, Combining accuracy and prior sensitivity for classifier design under prior uncertainty, Structural and Syntactic Pattern Recognition, Proc. SSPR2006, (Hong Kong, China), Lecture notes in computer science vol. 4109, Springer Verlag, Berlin, pages 512-521, August 2006
  • Landgrebe, T.C.W., Tax, D.M.J., Paclik, P., and Duin, R.P.W., The interaction between classification and reject performance for distance-based reject-option classifiers, Pattern Recognition Letters, Special issue on ROC analysis, pages 908-917, vol 27, number 8, June 2006
  • Landgrebe, T.C.W. and Duin, R.P.W, On Neyman-Pearson optimisation for multiclass classifiers, Sixteenth Annual Symposium of the Pattern Recognition Assoc. of South Africa, November 2005
  • Landgrebe, T.C.W., Paclik, P., Tax, D.M.J. and Duin, R.P.W., Optimising two-stage recognition systems, International workshop on multiple classifier systems, California USA, June 2005
  • Paclik, P., Landgrebe, T.C.W., Tax, D.M.J., and Duin, R.P.W., On deriving the second-stage training set for trainable combiners, International workshop on multiple classifier systems, California, USA, June 2005
  • Landgrebe, T.C.W., Paclik, P., Tax, D.M.J., Verzakov, S and Duin, R.P.W, Cost- based classifier evaluation for imbalanced problems, Structural and Syntactic Pattern Recognition, Proc. SSPR2004, (Lisbon, Portugal), pages 762-770, August 2004
  • Landgrebe, T.C.W., Tax, D.M.J., Paclik, P., and Duin, R.P.W and Andrew, C.M., A combining strategy for ill-defined problems, Fifteenth Annual Symposium of the Pattern Recognition Association of South Africa, pages 57-62, November 2004