KEDRI - All events
All events
8th June 2012 - 12th Conference on Neuro-Computing and Evolving Intelligence (NCEI'12)
NCEI’12 conference/workshop is a continuation of a previous conference series in NZ - ANNES, established in 1993. This conference aims at promoting research and development activities focused on advanced intelligent computing techniques by providing a vibrant and effective forum across a variety of disciplines. The scope of the conference includes a range of techniques such as Artificial Intelligence, Neural Networks, Evolutionary Computing, Brain Computer Interfaces, Informatics Theory and Applications, Computational Neuroscience, Neuromorphic computation, Bioinformatics, Pattern Recognition, Data Mining and Health Informatics.
The NCEI 2012 workshop is organized as a pre-conference workshop held just before the WCCI'12 in Brisbane, Australia (June 10-15, 2012).
Special session at the WCCI
Title: Spiking Neural Networks for Spatio- and Spectro-Temporal Data Modelling and Pattern Recognition: Methods, Systems, Applications
Date: 10 -15 June 2012
3 April 2012 - KEDRI Open Lecture on "Advanced Algorithms for Machine Learning and Remote Sensing Image Analysis" by Prof. Jie Yang
Advanced algorithms for machine learning are presented, such as Improved Feature Selection Algorithms for High-Dimensional Data; Improved Soft Subspace Clustering; Dimensionality Reduction by Manifold Learning. A system for remote sensing image analysis is presented, its application for ecological environment monitoring in Xinjiang Province, China is demonstrated.
Biography
Prof. Jie Yang received a bachelor degree in Automatic Control and a master degree in Pattern Recognition & Intelligent System in Shanghai Jiao Tong University. He received his Ph.D. at the Department of Computer Science, University of Hamburg, Germany in 1994. He is the Professor and Director of the Institute of Image Processing and Pattern recognition in Shanghai Jiao Tong University. He is the principal investigator of more than 30 nation and ministry scientific research projects in image processing, pattern recognition, data mining, and artificial intelligence, including two national 973 research plan projects, three national 863 research plan projects, three national nature foundation projects, five international cooperative projects with France, Korea, Japan, New Zealand. He has published more than 5 hundreds of articles in national or international academic journals and conferences.
3 March 2011 - KEDRI Seminar by Dr. Gisela Klette on "A Recursive Algorithm for the Computation of the Relative Convex Hull"
The relative convex hull of a simple polygon A, contained in a second simple polygon B, is known to be the minimum perimeter polygon (MPP).
Digital geometry studies a special case: A is the inner and B the outer polygon of a component in an image, and the MPP is called minimum length polygon (MLP). The MPP or MLP, or the relative convex hull, are uniquely defined. The paper recalls properties and algorithms related to the relative convex hull, and proposes a (recursive) algorithm for calculating the relative convex hull. The input may be simple polygons A and B in general, or inner and outer polygonal shapes in 2D digital imaging. The new algorithm is easy to understand, and is explained here for the general case. The worst case time complexity is quadratic, but it runs for "typical'' (as in image analysis) inputs in linear time.
Biography
Dr. Gisela Klette has a Master’s degree in Mathematics from the Friedrich-Schiller-University (Jena, Germany) and a PhD in medical image analysis from the University at Groningen (The Netherlands). She was working in a combined teaching and research position at the University of Auckland between 2002 and 2007. She is now a part-time lecturer at AUT.
Her research interests are in medical image analysis, in particular in applications of discrete geometry. She is a reviewer for journals such as IEEE Transactions in Pattern Analysis and Machine Intelligence, Theoretical Computer Science and Acta Cybernetica.
23 February 2010 – KEDRI Seminar on ”On the Puzzle of the Induction/Deduction Loop Bridging Perception and Symbolic Reasoning” and “Graphical Pattern Recognition: the LSRN Approach”
Speaker: Prof. Marco Gori
Title: On the Puzzle of the Induction/Deduction Loop Bridging Perception and Symbolic Reasoning
Speaker: Prof. Edmondo Trentin
Title: Graphical Pattern Recognition, the LSRN Approach
Prof. Marco Gori and Prof. Edmondo Trentin together with seminar participants
10 December 2009 – AUT Summer Graduation
On the AUT Summer Graduation day, three of KEDRI's PhD students had graduated - Anju Verma, Snjezana Soltic and Peter Hwang and three of KEDRI's Matsers students had also graduated with a First Class - Linda Liang, Sean Gordon and Peter Wang.
In addition to that, one KEDRI’s PhD student, Stefan Schliebs has received the Dean's Award for the Best Postgraduate Study in the School.
6 August 2009 – KEDRI Seminar on Kinetics models and laser photolysis
Title: Analysis of synaptic transmission and its regulation by channel
Speaker: Prof. Hiroshi Kojima, Laboratory for Cellular and Molecular Physiology, Tamagawa University, Tokyo Japan
The presentation will introduce the physiological experiments and techniques which enable us to understand the background of the synaptic transmission. Then it will explain how the rate constants of AMPA receptor channel kinetics model of neural cells which could induce the change in efficacy of synaptic transmission can be predicted with computer simulation.
Excitatory postsynaptic currents (EPSCs) were reconstructed by computer calculation. Moreover, electrical responses were measured from neurons by using laser uncaging experiments. It was shown that the properties of the evoked current responses by photolysis and those obtained spontaneous synaptic currents have the similar properties. It is also suggested that the present method based on kinetics models and uncaging method could be used for the investigation of the integration and regulation by neural cells.
On the photo from left to right: Prof. Nikola Kasabov (KEDRI), Prof. Hiroshi Kojima (Tamagawa University), Stefan Schliebs (KEDRI)
11th December 2008 - KEDRI seminar on "Progress and challenges in vision-based driver assistant systems"
Speaker: Professor Reinhard Klette, Computer Science Department, University of Auckland
The talk will discuss computer vision challenges in the context of vision-based driver assistant systems, certainly one of the most difficult areas of current 3D image analysis due to (1) the expectation of being "legally soundproof" at night or in the rain, at 120km/h, in Queen Street at rush hour, and so forth, also due to (2) the real time request, and (3) expectations to look about 3 seconds ahead - what may happen next in front of the car, at the location where the ego-vehicle is expected to be? Actually, car companies have started in 1995 to add vision solutions into their top-end models. What can we add to this in New Zealand, without having any car production line in the country? The talk will suggest collaboration between KEDRI/AUT an the .enpeda.. team.
21st and 28th February 2008 - KEDRI seminars on Technical characteristics and programming of the mobile robot "WITH"
Topic: Technical characteristics and programming of the mobile robot "WITH"
Presenter: Mr. Ryota Nishioka PhD student, Kyushu Institute of Technology (KIT), Japan
Supervisor: Prof. Takeshi Yamakawa
The speaker will present some technical characteristics, programming, details and instructions as to how to develop applications for the robot "WITH" built at the KIT. The robot is now available at KEDRI for joint projects based on the collaboration agreement between KEDRI/AUT and KIT.
Suggestions for new projects based on this robot should be sent to Dr. Paul (Shaoning) Pang: spang@aut.ac.nz
9th August 2007 - KEDRI Seminar
Speaker: Professor Ajit Narayanan,Head of School of Computing and Mathematical Sciences, AUT
Topic: Intelligent Bioinformatics and Cancer Systems Biology: The Computational Search for Killer Genes
The presentation will deal with a systems level view of cancer. Science goes through three interacting phases: measurement/observation, understanding, and control. Systems biology is the application of computational, mathematical and engineering concepts and techniques for understanding and, ultimately, controlling biological processes. The focus of the talk is on the following questions: (a) What are the basic structures and properties of a cancer biological network? (b) How does a cancer biological system behave over time under various conditions? (c) How does a cancer biological system maintain its robustness and stability? (d) How can we modify or construct cancer biological systems to achieve desired properties? The talk will start with an introduction to the 'standard' model of cancer before examining recent research from a cancer systems biology perspective that questions some aspects of the standard model, leading to new hypotheses of what causes cancer.
No previous knowledge of cancer or biology is assumed.
12th July 2007 - KEDRI OPEN DAY
KEDRI is organising an 'Open Day' to celebrate its achievements over the last five years, in the areas of research and postgraduate teaching.
Programme
General presentations:
- Introduction to KEDRI
- The KEDRI Brain-gene Ontology and Simulation System
- The KEDRI Computational Intelligence Repository
Centre presentations starting 1:30pm, in parallel, in KEDRI labs
Centre for Neurocomputing and Computational Intelligence
- Evolving Spiking Neural Networks
- Quantum inspired neurocomputing
- Evolving ontology-based machine learning environment
Centre for Neuroinformatics and Brain Study
- Computational neurogenetic modelling and case studies of learning, memory and epilepsy
- Integrated neuroinformatics--bioinformatics environment for research and discovery
Centre for Bioinformatics
- Gene expression profiling – SIFTWARE
- Gene regulatory network modelling – GeneNetXP
- Micro RNA discovery based on 2D image data analysis
Centre for Data Mining and Decision Support Systems
- NeuCom - Neuro-computing Decision Support Environment
- Renal function personalised medical decision support systems: GFR- ECOS and DOPPS
- Ontology-based personalised decision support system for chronic disease risk evaluation
Centre for Adaptive Pattern Recognition
- Aggregating intelligence for person authentication
- Evolving connectionist systems for robot control
- Modelling human motion and intelligent coaching systems for tennis and golf
Laboratory for English and Maori translation
- English and Maori on-line word translator
Contact Person: Prof. Nikola Kasabov, Email: nkasabov@aut.ac.nz
KEDRI Open-Day booklet is available for download as PDF here.
18th of April 2007 - Faculty of Design and Creative Technologies Research Forum Presentation by Prof. Nik Kasabov
Prof. Nik Kasabov presented a seminar at the faculty of design and creative technologies research forum. The topic was "The World of Information - Where Science, Art and Technology Meet".
Click here to download the presentation as a PDF file.
19th of October 2006 - KEDRI Seminar by Prof. Nik Kasabov
Topic: Brain, Gene and Quantum Inspired Computational Intelligence: Challenges and Opportunities
Presenter: Prof. Nikola Kasabov, FRSNZ, FNZCS, Director & Chief Scientist of KEDRI
The PDF version of the presentation can be downloaded here
24th of August 2006 - KEDRI Seminar by Dr. Petr Pancoska
Topic: Genes, Graphs and Proteins – Lessons for Novel Methods of Data Analysis
Biography
Dr. Petr Pancoska studied Chemistry at the Faculty of Sciences of the Charles University Prague while working simultaneously on research projects at the Institute of Organic Chemistry and Biochemistry of the Czechoslovak Academy of Sciences at Prague. He did his PhD. in experimental physics at the Faculty of Mathematics and Physics of the Charles University Prague. After postdoctoral work at the Bochum University (Germany) he was faculty at the Department of Chemical Physics of the Charles University Prague. He is also founding and board member of the International Centre for Discrete Mathematics, Theoretical Computer Science and Applications (DIMATIA) at the Charles University. From 1994-2003 he was faculty at the Department of Chemistry of the University of Illinois at Chicago. Currently he is faculty at the Stony Brook University. Research interests cover applications of advanced data analysis methods, novel applications of mathematics in bioinformatics, cancer research, bio-nanotechnology, molecular biology, virology, drug design and disease diagnostics, optical spectroscopy of biomolecules, experimental and theoretical studies of primary processes of photosynthesis.
Abstract
The evolutionary history of nucleotide and protein sequences is usually inferred from multiple sequence alignments. We show in the first part of the presentation that – using an innovative graph-theoretical approach – it is possible to extract evolutionary constraints within single sequences. This novel approach has firm physico-chemical basis: The multiplicity of realizations of the “wild type” thermodynamic state of any gene position is calculated via representation of gene sequence via Eulerian graph. This innovative descriptor reveals previously unseen dimension of information content within genomes. The multiplicity technique is entirely model-free, there is no need to do multiple sequence alignments or parameterize evolutionary rates or distributions. A direct observation of the sequence via thermodynamic “tolerance” reveals where evolution has happened and how important mutations in each particular position are in comparison to other positions (even if they have the same local sequence). No other technique (alignment entropy, codon bias, Ka/Ks ratio) has a similar potential. We show practical applications of this novel analysis in drug design, prediction of emergence of mutations of influenza viruses that induce drug resistance, in quantitative prognostics of treatment of HIV patients and for identification of “barcode” segments of pathogens for microarray-based bio-defense applications.
In the second part we will use the insight gained from the above biological applications for more general analysis of large graphs and networks. First we will exploit the unique topological structure of any Eulerian graph E. We show that a (weighted) graph g can be converted into an (oriented) multigraph G. Then : i) Each E-subgraph of G can be decomposed into a (small)“basis”of independent cycles and encoded digitally in a vector; ii) E-subgraph of a network is deterministically associated with a linear sequence of events. Moreover, the multiplicity of this sequence of events provide a measure of its uniqueness; iii) Topology of any E-subgraph defines the “command structure” of interacting blocks, which govern the events sequence. iv) The results of i)-iii) can be used for direct design of networks with pre-defined properties and interactions between its components.
Next we will exploit the broader consequences of the fact that all biological “information media” are chiral. We will show some results of the theoretical analyses of this natural phenomenon for chiral subgraphs C of g. i) Ruch’s theory associates the topology of C (via characters of its symmetry and permutation groups) with a unique chirality function that describes the mode of interactions between the C components. ii) The existence of parity-odd and parity even observables can represent two different responses of a network to a stimulus. We will demonstrate on a simple conceptual example, that the magnitude of the differential response depends on a level of interaction between the components of C.
3rd of August 2006 - KEDRI Seminar by Prof. Mario Fedrizzi
Topic: Modelling Consensual Dynamics in Group Decision Making in the Framework of Fuzzy Preferences
Biography
Prof. Mario Fedrizzi is a Professor and Chair of Mathematics for Decisions, Faculty of Ecomomics, and the Deputy Rector, University of Trento, Italy. His current research focuses on: utility and risk theory; stochastic dominance; group decision analysis; decision support systems; fuzzy decision analysis; fuzzy mathematical programming; fuzzy regression analysis; consensus modeling in uncertain environments. Prof. Fedrizzi is also involved in consulting activities for information systems and DSS design and implementation; office automation; process quality control; project management; expert systems and neural nets in financial planning; ERP systems design and implementation.
Abstract
One of the central issues to address when we are faced with the preference aggregation task in social choice theory is that of modelling consensus. The notion of consensus plays an important role whenever the social choice scheme is based on a strategy of gradually combining the various individual preferences into some form of collective preference structure.
In the development of fuzzy approaches to group decision making the notion of consensus has evolved from a binary index of unanimous agreement to a graded (soft) measure of collective agreement among the individual decision makers whose preferences are represented by fuzzy binary relations.
Here we propose some extensions of the fuzzy approaches to consensus reaching processes combining a soft measure of collective dissensus with an inertial mechanism of opinion changing aversion. The collective consensual trend corresponds to a process of anisotropic diffusion among the individual preferences structure. The anisotropy is designed so as to outline and enhance the natural group segmentation into homogeneous subgroups. The individual inertial mechanism, on the other hand, opposes changes from the original preferences and provides an appropriate framework to deal with preference outliers.
1st of August 2006 - KEDRI Open Lecture Series X
Topic: Theories of Decision Making Under Risk and Uncertainty: A Selected Survey of Mathematically-based Approaches in the Second Half of the 20th Century
Biography
Prof. Mario Fedrizzi is a Professor and Chair of Mathematics for Decisions, Faculty of Ecomomics, and the Deputy Rector, University of Trento, Italy. His current research focuses on: utility and risk theory; stochastic dominance; group decision analysis; decision support systems; fuzzy decision analysis; fuzzy mathematical programming; fuzzy regression analysis; consensus modeling in uncertain environments. Prof. Fedrizzi is also involved in consulting activities for information systems and DSS design and implementation; office automation; process quality control; project management; expert systems and neural nets in financial planning; ERP systems design and implementation.
Abstract
The monumental treatise on rational choice and the theory of games of John von Neumann and Oskar Morgenstern, published in 1947, have profoundly influenced mathematical research in decision making modelling through the rest of the 20th century. K. Arrow’s approach on social choice theory, L. Savage’s axiomatic foundations for subjective utility theory, G. Debreu’s theory of value and R. Keeney and H. Raiffa’s work on preferences and value tradeoffs have been directly derived from the seminal work of von Neumann and Morgenstern. The decision making framework introduced by the two authors, usually referred to as the expected utility model, have come under increasing fire due the weakness of its explanatory power for real world decision processes. P. Fishburn, D. Kahneman and A Tversky, M. Machina, D. Schmeidler and others introduced generalizations accommodating reasonable behaviours of decision makers that were considered inconsistent with the axiomatic framework of the original expected utility model.
The main scope of the lecture is to provide an historical overview of the basic normative decision making models based on the representation of preference structures. Strengths and weaknesses of the different approaches will be enlightened in the perspective of the impact of the models on the ways in which decisions are made. Some basic problems related to the representation and management of risk, vagueness and imprecision will be addressed.
The PowerPoint presentation of the lecture can be downloaded here.
8th of June 2006 - KEDRI Seminar by Dr. Svetlana Shevenko
Dr. Shevenko will be speaking on Automatic processing Japanese texts, Automatic recognition Chinese characters and Application of the research results for developing accelerated language courses.4th of May 2006 - KEDRI Open Lecture Series IX
Topic: The Comparison of the Spatio-temporal Learning Rule and the Hebbian Learning Rule
Speaker: Prof. Minoru Tsukada,
Biography
Prof. Minoru Tsukada is a Professor in the Department of Intelligent Information Systems, Faculty of Engineering, Tamagawa University and he is also the General Manager of Research Institute, Tamagawa University
Mailing address: 6-1-1, Tamagawa-gakuen, Machida, Tokyo, 194-8160, Japan,
Web: Professor Tsukada Minoru profile
E-mail: tsukada@eng.tamagawa.ac.jp
Phone/Fax: +81-(0)42-739-8430
Field of Specialisation: Computational Neuroscience and Modeling of Neural Networks, Intelligent Informatics
Research topics
A spatio-temporal learning rule and information representation in memory networks; Self-organisation of information representation; A dynamic aspect of neural coding.
To experimentally investigate self-organization of information representation in the neural memory system, and to design a theoretical model of the hippocampal-cortical memory system of storing and retrieving information.
The following kinds of information representation are identified in experiment and model study: Dynamic cell assembly; Information integration by temporal correlation among events; A spatio-temporal learning rule; A dynamic non-linear process of storing and retrieving information
Abstract
The Hebbian synaptic learning rule requires coactivity of presynaptic and postsynaptic neurons. However, under some conditions, information regarding the postsynaptic action potentials, carried by backpropagating action potentials, can be strongly degraded before it reaches the distal dendritic synapse of the hippocampal CA1 (Spruston, et al.,1995; Andreasen and Ross,1995; Callaway and Ross,1995; Stuart, et al.,1997; Golding et al., 2001). Yet, recent results (Golding et al., 2002) have shown that LTP can indeed occur at synapses on distal dendrites of hippocampal CA1 pyramidal neurons, even in the absence of a postsynaptic somatic spike.
Based on results observed in hippocampal LTP, induced by various spatiotemporal pattern stimuli (Tsukada et al.,1990,1994), the spatiotemporal learning rule (STLR) was proposed by Tsukada et al. (1996,1998). The novel point of this learning rule was “Cooperative plasticity without a postsynaptic spike”, which we tested physiologically in the CA1 hippocampal network (Yamazaki et al.,2006), and its temporal summation.
The learning rule incorporated two dynamic process: fast (10 to 30ms) and slow (150 to 250ms). The fast process works as a time window to detect a spatial coincidence among various inputs projected to a weight space of the hippocampal CA1 dendrites, while the slow process works as a temporal integrator of a sequence of events.
In a previous paper (T.Aihara et al.,2000) the decay constant of fast dynamics was identified as 17ms by parameter fitting to the physiological data of LTP, Cell assemblies were synchronized at this time scale, while matches the period of the hippocampal gamma oscillation, and that of the slow is 169ms, which corresponds to a theta rhythm. We (Tsukada and Pan, 2005) systematically examine the functional difference between STLR and Hebbian learning rules in a single–layered neural network, computing their ability to differentiate spatiotemporal sequence. In this paper, we tested physiologically the cooperative plasticity without a postsynaptic spike in the CA1 hippocampal network.
17th of November 2005 - KEDRI Open Lecture Series VIII
Speaker: Prof. L. T. Koczy
Topic: Recent Advances in Fuzzy Model Identification
Biography
Laszlo T. Koczy M. Sc. E. E. (1975); M. Phil. Control E. (1976); Ph.D. (1977), Technical University of Budapest, Doctor of the Hungarian Academy of Science, (1998, the highest earnable postdoctoral degree in Hungary) Visiting positions: Australian National University (Canberra); Murdoch University (Perth, Australia), University of New South Wales (Sydney, Australia); J. Kepler Universitat Linz (Austria); University of Trento (Italy); Tokyo Institute of Technology (Yokohama, Japan; Chair Professor in 1993/94); Pohang Institute of Science and Technology (Korea). Summer University lecturing: Helsinki University of Economics, University of Helsinki (Finland), University of Minas Gerais (Belo Horizonte, Brazil), Dalian Maritime University (China). Research interests: Telecommunication systems, Intelligent models and systems, Very large and complex systems and networks. Professional Societies: International Fuzzy Systems Association: Past President 2003-2001; President 2001-2003; President Elect, 1999-2001; Vice President, 1995-99. IEEE: Senior Member, 1999-; NNS Regional Activities Chair 2001-2002; Chapters Chair 2002-2003, Administrative Committee member, 2004- .EURO WG on Fuzzy Sets (currently EUSFLAT): Founding Member, 1975-. Hungarian Fuzzy Society: Founding President 1990-1999, now Life Honorary President. He has published 326 papers including some textbooks.
Abstract
Automatic fuzzy rule based model identification started with Sugeno and Yasukawa's paper in 1992. The first approaches were based on clustering, especially fuzzy c-means clustering. Later several essential improvements have been added and alternative learning techniques were applied (with T. Gedeon and A. Chong). Bacterial algorithms introduced by Furuhashi could be used for slow but global model optimisation (with J. Botzheim and B. Hamori), while the Levenberg-Marquardt algorithm successfully used for neural network optimisation could be extended to more complex fuzzy models as a fast but locally converging technique (with A. Ruano, C. Cabrita and J. Botzheim). Our most recent results combine the advantages of these two, in the form of a Bacterial Memetic Algorithm, where fast and global optimum could be reached. Some benchmark problems will illustrate the power of the new approach.
30th of June 2005 - KEDRI Open Lecture Series VII
Speaker: Prof. J.J. Wright
Topic: Dynamics and Organization of the Cerebral Cortex
Biography
Jim Wright is a psychiatrist, and an Honorary Professor in the Faculty of Medicine and Health Sciences. He took his medical degree at the University of Otago, and was subsequently educated at the California Institute of Technology and the University of London, before returning to Auckland. His primary research interest is, and has always been, the dynamics of the brain. He was at one time head of the Department of Psychiatry and Behavioural Science at the Medical School, and more recently held a personal chair at the Mental Health Research Institute of Victoria. In 2003 he shared the Royal Societies of Australia Prize for Interdisciplinary Research, with physicists and physiologists of Sydney University. He currently is engaged in ongoing research in Brain Dynamics, with colleagues in Auckland, Hamilton, Sydney and Melbourne.
For further details please see flyer attached.
13th December 2004 - KEDRI Open Lecture Series VI
“A Neurobiological Theory on the Meaning of Information” by Prof. Walter Freeman.
Prof. Walter J Freeman is currently a professor of the Graduate School, U.C. Berkeley from 1994. From 1967 to 1994, he was a professor of Physiology, University of California, Berkeley. He was a president of International Neural Network Society and a president of the Pavlovian Society. He was awarded; Francis Perkins Prize for first place in basic science (1952), Oliver P. Douglas Award (1958), A.E. Bennett Award (1964), John Simon Guggenheim Fellowship Award (1965), MERIT Award (1991), Pioneer Award of Neural Networks Council, IEEE (1992), and so on. He is a Fellow of IEEE.
26th March 2004 - KEDRI Open Lecture Series V
Miklos N. Szilagyi, is a Professor of Electrical and Computer Engineering at the University of Arizona. His major research interests are in: Computer Simulation, Neural Networks, Artificial Intelligence, Particle Beams and Optics, Physical Electronics, Electromagnetics, Computer-Aided Design, Biomedical Engineering, Applied Physics, Applied Mathematics, Science and Education Policy, Management and Administration.
The seminar will include the following topics:
Various approaches (mathematics, AI, neural networks, genetic algorithms) What is agent-based simulation? How to build a model? How to implement it? Example 1. Sugarscape, Example 2. Dilemma Resources for further study.
