Dr Joab Winkler

PhD

Department of Computer Science

Reader

MSc Admissions Tutor

City College Liaison

Member of the Machine Learning research group

Joab Winkler profile photo
Profile picture of Joab Winkler profile photo
j.r.winkler@sheffield.ac.uk
+44 114 222 1834

Full contact details

Dr Joab Winkler
Department of Computer Science
Regent Court (DCS)
211 Portobello
Sheffield
S1 4DP
Profile

Joab Winkler is a Reader in The Department of Computer Science at The University of Sheffield. He obtained his undergraduate and PhD degrees at Imperial College London and University College London, respectively. He worked for a few years in industry, before returning to university to conduct research into algebraic and numerical properties of curves and surfaces in computer-aided design systems. He has extended this interest in numerical methods to work on blind image deconvolution, optical flow and neural networks. More details are in the section 'Research interests'.

Research interests

Joab Winkler’s main research interest is the algebraic and numerical properties of curves and surfaces in computer-aided design systems. Most of this work has been performed using resultant matrices, and this has led him to consider more general issues of robust computations on polynomials that are corrupted by added noise. Examples include the computation of a structured low rank approximation of the Sylvester resultant matrix, the deconvolution of two polynomials and the determination of an approximate greatest common divisor of two polynomials.

He has developed a polynomial root solver for the determination of multiple roots of the theoretically exact form of a polynomial, when the coefficients of the given polynomial are corrupted by added noise. He has applied the methods used in this work on polynomials to the deblurring of an image when the point spread function is unknown and must therefore be computed. This is a difficult problem because it is necessary to determine the rank of a matrix whose entries are subject to error. This is one of the most difficult problems in linear algebra.

He has also considered numerical and computational issues in optical flow, which is an important problem in computer vision, in order to identify problems that arise in two commonly used algorithms. 

More recently, he has started work on numerical properties of echo state networks, which are one form of a neural network. 

Publications

Books

  • Winkler J, Lawrence N & Niranjan M (2005) Lecture Notes in Computer Science (Including Subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics): Preface. RIS download Bibtex download

Journal articles

Conference proceedings papers

  • Alkhaldi N & Winkler JR (2016) IEEE International Conference on Image Processing (ICIP 2015). IEEE International Conference on Image Prcoessing (pp 784-788). Quebec City, Canada, 27 September 2015 - 30 September 2015. View this article in WRRO RIS download Bibtex download
  • Alkhaldi N & Winkler J (2015) Blind image deconvolution using the Sylvester resultant matrix. 2015 IEEE International Conference on Image Processing (ICIP), 27 September 2015 - 30 September 2015. RIS download Bibtex download
  • Winkler JR (2015) The Sylvester Resultant Matrix and Image Deblurring. 8th International Conference on Curves and Surfaces (pp 461-490). Switzerland, 12 June 2014 - 18 June 2014. RIS download Bibtex download
  • Winkler JR (2015) The Sylvester Resultant Matrix and Image Deblurring (pp 461-490) RIS download Bibtex download
  • Winkler JR, Martin, R & Sabin, M (2007) Mathematics of Surfaces XII. Springer (LNCS 4647), Vol. 4647. Sheffield, 4 September 2007 - 7 September 2007. RIS download Bibtex download
  • Winkler JR (2007) Condition numbers and least squares regression. Mathematics of Surfaces XII, Proceedings, Vol. 4647 (pp 480-493) RIS download Bibtex download
  • Winkler JR (2006) Structure Preserving methods for the computation of approximate GCDs of Berstein polynomials. 6th International Conference on Curves and Surfaces (pp 11-20). Avignon, France, 29 June 2006 - 5 July 2006. RIS download Bibtex download
  • Samani A, Winkler J & Niranjan M (2006) Automatic face recognition using stereo images. 2006 IEEE INTERNATIONAL CONFERENCE ON ACOUSTICS, SPEECH AND SIGNAL PROCESSING, VOLS 1-13 (pp 5771-5774) RIS download Bibtex download
  • (2005) Deterministic and Statistical Methods in Machine Learning, First International Workshop, Sheffield, UK, September 7-10, 2004, Revised Lectures. Deterministic and Statistical Methods in Machine Learning, Vol. 3635 RIS download Bibtex download
  • Winkler JR (2005) A comparison of condition numbers for the full rank least squares problem. DETERMINISTIC AND STATISTICAL METHODS IN MACHINE LEARNING, Vol. 3635 (pp 296-318) RIS download Bibtex download
  • Winkler JR (2005) Numerical and algebraic properties of Bernstein basis resultant matrices. COMPUTATIONAL METHODS FOR ALGEBRAIC SPLINE SURFACES (pp 107-118) RIS download Bibtex download
  • Winkler JR & Goldman RN (2003) The Sylvester resultant matrix for Bernstein polynomials. Curve and Surface Fitting: Saint-Malo 2002 (pp 407-416). Saint-Malo, 2002 - 2002. RIS download Bibtex download
  • Winkler JR & Niranjan, M (2002) Uncertainty in Geometric Computations. Uncertainty in Geometric Computations. Sheffield, 5 July 2001 - 6 July 2001. RIS download Bibtex download
  • Winkler JR (2002) Properties of the companion matrix resultant for Bernstein polynomials. UNCERTAINTY IN GEOMETRIC COMPUTATIONS, Vol. 704 (pp 185-198) RIS download Bibtex download
  • Winkler JR (2001) Computational experiments with resultants for scaled Bernstein polynomials. MATHEMATICAL METHODS FOR CURVES AND SURFACES: OSLO 2000 (pp 535-544) RIS download Bibtex download
  • Winkler JR & Ragozin DL (2000) A class of Bernstein polynomials that satisfy Descartes' rule of signs exactly. MATHEMATICS OF SURFACES IX (pp 424-437) RIS download Bibtex download
  • Winkler JR () The sensitivity of linear algebraic equations. 5th SIAM Conference on Applied Linear Algebra. Snowbird, Utah, USA, 15 June 1994 - 18 June 1994. RIS download Bibtex download
  • Winkler JR () Numerically stable conversion between the Bezier and B-spline forms of a curve. Curves and Surfaces with Applications in CAGD (pp 465-472). Chamonix, 30 June 1997. RIS download Bibtex download
  • Winkler JR () The numerical condition of roots of polynomials in Bernstein form. Proceedings of the 8th IMA Conference on The Mathematics of Surfaces (pp 265-280). Birmingham, 31 August 1998 - 2 September 1998. RIS download Bibtex download

Preprints

Grants

Research Grants

  • Sparse linear models: Their existence and stability, EPSRC, 09/2022 - 07/2023, £79,384, as PI
  • Travel Grant, EPSRC, 07/2001 to 08/2001, £1,430, as PI
  • Tensor Tomography for the Three Dimensional Photoelasticity, EPSRC, 11/2002 to 04/2006, £38,818, as PI
  • Robust Computations in Geometric Modelling, EPSRC, 02/2005 to 01/2008, £71,785, as PI
  • Travel Grant, EPSRC, 10/2005 to 01/2006, £4,100, as PI
  • Polynomials and geometric modelling, ROYAL ACADEMY OF ENGINEERING (THE), 01/2010 to 12/2012, £23,613, as PI
Professional activities and memberships
  • Awarded an EPSRC Advanced Research Fellowship (1995)
  • Co-organiser of a workshop, supported by the EPSRC, on The Representation and Management of Uncertainty in Geometric Computations (2001)
  • Co-organiser of The Sheffield Machine Learning Workshop, supported by the EPSRC (2004)
  • Organised the Summer School, supported by the EPSRC, Solving Polynomial Equations and Structured Matrix Methods for Approximate GCD Computations (2007)
  • Awarded a Global Research Award by The Royal Academy of Engineering (2010-2011)
  • Co-organiser of a conference in Kalamata, Greece on structured methods in numerical linear and multilinear algebra (2014)