Dr Elżbieta Pękalska  
EPSRC research fellow

pekalska [at] cs.man.ac.uk
Kilburn building
School of Computer Science
University of Manchester
Oxford Road
M13 9PL Manchester, UK
tel: +44 (0)161 275 4763

Some information on Ela Pękalska can be found in her CV.

Publications
Google Scholar: Pekalska

Key values

1. Wisdom: Spirit, intelligence, consciousness, knowledge, understanding, clarity, elegance, holistic view, integration, simple solutions.
2. Love: warmth, compassion, caring, enthusiasm, empathy, acceptance, growth.
3. Reaching the core: exploration, discovery, pursuit into unknown, courage, persistence, truth.
4. Contribution: care, value, service, honesty, impact, making a difference.
5. Synergy: intuition, creativity, cooperation, solution, flow.
6. Joy: fun, curiosity, passion, zeal, fulfilment, humour.
7. Professionalism: integrity, competency, efficiency, resourcefulness, virtue.
8. Expression: art, movement, dance, harmony, elegance, freedom, nature.
9. Recognition: similarity, understanding, appreciation, co-creation, partnership.

Book on generalised kernel representations
Are you interested in generalised kernel methods? Then you may like to have a look at this book. Dissimilarity representations are generalisations of the traditional kernels. In these pairwise representations, every value describes proximity between pairs of objects. Any metric, non-metric or indefinite measure can be used, provided that it is meaningful for the problem. An extension leads to conceptual proximity representations which encodes proximity between an object and a model (of a class or a cluster). This notion leads further to stacking and trained combiners. The book starts with theory and then follows to practice. Dissimilarity representations are applied to a range of problems in pattern recognition. These include visualisation, data exploration, complexity of representation, clustering, classification, novelty detection and combining.

Elżbieta Pękalska and Robert P.W. Duin The Dissimilarity Representation for Pattern Recognition. Foundations and Applications. World Scientific, Singapore, December 2005. Book: 636 pages. ISBN: 981-256-530-2. Buy: Amazon UK, Amazon USA, or Amazon Germany. Explore: Google-Books.

         TOC
         Chapter 1: Introduction

Related software
PRTools, Pattern Recognition Toolbox in Matlab.

Current involvement in projects

• EPSRC fellowship grant: "Statistical learning with general proximity measures".

Previous involvement in projects

• EU FET project within the FP7 framework: SIMBAD - "Beyond features: Similarity-based Pattern Analysis and Recognition". The project involves Delft Univerisity of Technology (DUT), ETH Zurich, IST Lisbon, University of Venice, University of Verona, and University of York. I work with the DUT side. The overall aim is to develop new methods for pattern analysis and machine learning based on potentially non-(geo)metric similarity data.
• ARC British Council grant for the British-German collaboration: "Indefinite Kernel Methods and Learning in Proximity Spaces".

Interests
My interests include pattern recognition (PR), machine learning (ML), image analysis, artificial intelligence, neuroscience and psychology. My work is mostly in the fields of PR and ML, with the emphasis on representation and generalisation. My key questions refer to the issue of representation, generic combining paradigms and the meaning of proximity in the learning from examples. I am also interested in various learning scenarios, such as novelty detection, transductive learning, active learning, semi-supervised learning, learning with data that disobey iid requirements, collaborative filtering, etc. I am keen both to explore and to understand Bayesian approaches, Gaussian processes, MDL, embeddings, (non-)linear mappings, indefinite kernels and the theory of Banach, Hilbert and Krein spaces.

Would you like to learn Matlab?
Personally, I find MatLab a great tool. I once wrote a Matlab manual for the students of physics. The idea was to teach them some concepts of programming by using Matlab. Maybe you like it as well. Click here (67 pages).

Collaborators

• Dr Manuele Bicego, University of Verona, Italy
• Dr Gavin Brown, University of Manchester, UK
• Prof. Horst Bunke,University of Bern, Switzerland
• Dr Robert Duin, Delft University of Technology, The Netherlands
• Prof. Ana Fred,Technical University of Lisbon, Portugal
• Dr Bernard Haasdonk, University of Stuttgart, Germany
• Prof Ludmila Kuncheva, University of Wales, UK
• Dr Pavel Paclik, Delft University of Technology, The Netherlands
• Dr David Tax, Delft University of Technology, The Netherlands

Additional activities
I am in the Program Committee of KDD-2011, MCS-2011, ICANN-2011, HAIS-2011, SIMBAD-2011 and ICPRAM-2012. Luca Cazzanti and I organized a special session on Learning from pairwise relationships at CIP-2010.

Reviewer for the journals: JBI, PAA, PR, PRL, TPAMI, TSMC, TNN, TSP, JMLR.
Program committee member for: ICPR, MCS, S+SSPR, PRIS, ICANN, CAIP, CIP, HAIS, ICML and NIPS.

Awards
2010: Classifier Domain of Competence Contest at ICPR-2010, Champion of the Neighborhood Dominance Test + award for offline training; Contest paper. Joint work with Bob Duin, Marco Loog and David Tax.

2008: Best Paper Award for the "Gesellschaft fuer Klassifikation", 2008 (bestowed in January 2009): Classification with Kernel Mahalanobis Distance Classifiers, joint work with Bernard Haasdonk. Published at the German Classification Society Annual Conference, 2008.

2008: Pattern Recognition Society Best Paper Award for 2006 (bestowed in November 2008): Prototype Selection for Dissimilarity-based Classifiers, joint work with Pavel Paclik and Robert Duin. Published in Pattern Recognition, vol. 39, issue 2, 189-208, 2006.

2007: DAAD-ARC research grant.

2006: EPSRC fellowship.

2005: Award for the Best Presentation at International Conference on Computer Recognition Systems, funded by the Association for Image Processing in Poland.

Mathematical genealogy
My descendant line includes some interesting people:

1. Erhard Weigel (PhD, 1650), advisor of
2. Gottfried Wilhelm Leibniz (PhD, 1666), advisor of
3. Johann Bernoulli (PhD, 1694), advisor of
4. Leonhard Euler (PhD, 1726), advisor of
5. Johann Friedrich Hennert (PhD, ?), advisor of
6. Jan Hendrik van Swinden (PhD, 1766), advisor of
7. Antonius Chaudoir (PhD, 1773), advisor of
8. Cornelius Ekama (PhD, 1794), advisor of
9. Pieter Johannes Uylenbroek (PhD, 1822), advisor of
10. Petrus Leonardus Rijke (PhD, 1836), advisor of
11. Hendrik Antoon Lorentz (PhD, 1875), advisor of
12. Leonard Salomon Ornstein (PhD, 1908), advisor of
13. Hendrik Berend Dorgelo (PhD, 1924), advisor of
14. Cornelis Johannes Dionisius Maria Verhagen (PhD, 1942), advisor of
15. Robert Pieter Wilhelm Duin (PhD, 1978), together with Ian Theodore Young advisors of
16. Elżbieta Pękalska (PhD, 2005)

Passion: Wisdom and Understanding

• External: Science
• Internal: Personal development. Coaching.
• Pattern recognition and conscious living blog.