Informatics Institute Annual Report 004 Chapter introduction

Informatics Institute
Annual Report 2004


Chapter 1


INTRODUCTION
The research in the Informatics Institute focuses on interactive complex information systems, processing large amounts of data in a distributed setting. This research goal is reflected in the three laboratories of the institute: the Computing, System Architecture, and Programming Laboratory (CSP), with a focus on Complex systems, Chair prof. Peter Sloot, the Intelligent Systems Laboratory Amsterdam (ISLA), with a focus on complex data- and information structures, Chair prof. Arnold Smeulders, and the Human Computer Studies (HCS) laboratory, with a focus on interaction in complex systems, Chair prof. Bob Wielinga.
In 2004 the Human Computer Studies laboratory (HCS) was founded as a merger of the Social Scientific Informatics group joining us, from the Faculty of Social and Behavioral Sciences, with the information management and the machine learning groups of our institute.
The year 2004 was a very good year for the Institute. The result of year-long investments materialized. The results of the national quality assessment of computer science made clear that the institute is in good shape. The overall scores of our institute were very good indeed. The ISIS group of prof. Arnold Smeulders in ISLA was awarded the highest rank of all groups in the Netherlands. Other groups of the institute were among the best as well.
The year 2004 was also the year in which two bsik-projects initiated from the UvA have been awarded. They are VL-e and MultimediaN, of which you will read more in the report. In addition, the institute participates in a third bsik-project, ICIS. And, in 2004, two new chairs were appointed in our institute; Prof Chris Jesshope, the future successor of prof. Bob Hertzberger, on the chair of system architecture engineering in the CSP-laboratory, and prof. Maarten de Rijke on the new chair, in ISLA, of Information Processing and Internet.The first ever VENI-award was awarded to dr. J-M. Geusebroek, ISLA, on the theme of cognitive vision.
The members of the Institute deliver the majority of courses in the curriculums of Computer Science, Artificial Intelligence and Information Systems. The new Bachelor-Masters structure has required a major effort, by the members of the institute, in restructuring many of the individual classes, a task they dealt with aptly.
The institute currently holds 8 full professorships, 7 part-time professorships and 9 associate professorships. In the institute there were 20 assistant professors, 37 postdocs, 48 PhD-students, 22 software developers and 1 lecturer. Many PhD-students and postdocs were funded by the national science foundation, the royal academy, public-private funding, and industry. Overall, external funding covers some 50% of our total research.
In 2004, we enjoyed 7 PhD-graduations on diverse topics ranging from An agent based architecture for constructing Interactive Simulation Systems and Style characterization of machine printed texts.

The staff members of the institute have generated 61 journal publications and 201 publications in refereed conference proceedings, a total of 262 publications.

The institute has made an effort to popularize computer science by displaying the fun and results of the institute in television and radio programs such as covering the topic of road traffic simulation in Vara program “Nieuwslicht” by prof. Peter Sloot.
The Institute participates in three national research schools: ASCI, the research school on Architecture and Image Systems (CSP and ISLA), IPA the research school for Programming and Algorithms (CSP) and SIKS the research school on Artificial Intelligence and Knowledge Based Systems (HCS and ISLA).
All in all 2004 was a good year for the institute with many promising developments for the future.
Prof.dr.ir. Frans C.A. Groen, Director

Chapter 2 Fundamental Research

The Laboratory for Computing, System architecture, and Programming
The Computing, System architecture, and Programming Laboratory (CSP-Lab), integrates three core computer science research questions: How to build complex computer systems, how to program them and how to process information on them. The laboratory consists of the groups on Parallel Systems Architecture, Programming Methods and Environments and Computational Informatics.
The Parallel Systems Architecture program involves the design and evaluation of parallel hardware and software architectures as integrated systems with emphasis on experiments. The project includes applications of distributed architectures.
The Programming Methods and Programming Environments program focuses on the one hand on generating programming environments given a formal language definition, in particular the construction of generic user-interfaces and the development of generic methods for the textual and graphical representation of structured objects. On the other hand the program focuses on development of a process theory and tools that can be used to specify and verify concurrent communicating or programmed systems.
The program Computational Informatics focuses on how global macroscopic processes emerge from local microscopic rules and interactions and is inspired by the paradigm of complex systems. Through modelling and simulation they seek to discover the way information is being processed in such systems. High dimensional cellular automata are studied as a representation of the computation. The applicability of this approach is validated through the development of high performance problem solving environments for asynchronous natural processes.

1. Computational Informatics
General information
Contact person : Prof Dr P.M.A. Sloot

Telephone : +31 - 20 - 525 7462

URL : http://www.science.uva.nl/research/scs/

Fax : +31 - 20 - 525 7419

E-mail : [e-mail]
Position within the Organisation
"Computational Informatics" is one of the three research groups of the Computing, System Architecture and Programming Laboratory of the Informatics Institute of the Faculty of Science.
The Computational Informatics group closely collaborates with other groups within the University, especially in Mathematics and the Natural Sciences, and is an important player in the Amsterdam Centre for Computational Science (ACCS). (See http://www.science.uva.nl/research/accs/) The Section Computational Science Participates in the Dutch graduate research school `ASCI'.
Characterization
Nature has a very efficient way to process information. Most natural processes are inherently parallel. We seek to discover, describe and use the way information is being processed in natural systems. We adopt the notion of complex systems to study how (global) macroscopic processes emerge from (local) microscopic rules and interactions. High dimensional cellular automata are used as a representation of these computational processes. The applicability of this approach is validated through realistic, high performance simulations of natural processes.
Key Words
Problem Solving Environments and Metacomputing; Distributed and Grid Computing; Interactive Algorithms; Modelling and Simulation of Complex Systems; Parallel Discrete Event Simulation, Self-organized criticality.
NOAG-I -Themes Modelling, Simulation and Visualization; Parallel and Distributed Computing.
Main themes
Our research on how global macroscopic processes emerge from local microscopic rules and interactions is inspired by the paradigm of complex systems. Through modeling and simulation we seek to discover the way information is being processed in such systems. High dimensional cellular automata are used as a representation of the computation. The applicability of this approach is validated through the development of high performance problem solving environments for asynchronous natural processes.
Modelling and Simulation
Nature is inherently asynchronous: Hubermann, PNAS, 1997
Modelling and simulation requires mapping of (natural) phenomena onto a model, mapping the model onto a parallel solver, and finally mapping the parallel solver to the computing resource. We do not follow the conventional partial differential equations approach in modelling, rather we try to model by directly mapping natural phenomena on (intrinsically parallel) solvers. For this we adopt the concept of Cellular Automata (CA). CA are spatio-temporal discrete, decentralized, finite state systems consisting of a large number of simple identical components with local connectivity. The research focuses on the information processing in asynchronous CA's and their execution behaviour on distributed systems.
Problem Solving Environments and Visualization
Progress in natural science comes from taking things apart, progress in computer science comes from bringing things together: W.D. Hillis, MIT, 1982.
We strongly believe in experimental validation of theoretical work. We investigate the mapping of Cellular Automata (CA) to parallel, distributed and hybrid architectures. Both architectural design and parallelisation of CA-models for these designs guide our research. The usability, efficiency, stability, and the correctness of our models are tested through their use in interactive Problem Solving environments. We adopt the common software-component architecture approach to obtain inter operable, modular PSE's that integrate modelling, simulation, interaction and visualization on Grid-like distributed compute systems.
2004 Results
Introduction

Within each of the two main themes in our research, a number of projects are being pursued. Sometimes these projects intersect both themes. We will start this section with the one thesis defended in 2004, by Z. Zhao, and then move on to a thematic report on the projects that were active in 2004. 2004 also saw the publication of the assessment of the Computer Science research institutes in the Netherlands, performed by a team of scientists of international renown on behalf of the Quality Assurance Netherlands Universities (QANU). In this report our research received the qualification “very good” on all scores.

Theses

After the four theses defended in 2003, 2004 saw only a single thesis defence, by Z. Zhao. In his thesis “An agent based architecture for constructing Interactive Simulation Systems” he describes an architecture for ISS that separates the control of the integrated from low-level communication infrastructure. In this manner existing components can be flexibly reused, without repeated reprogramming or restructuring of the basic software. The system can be seen as a system supporting a flexibly reconfigurable workflow between the constituent components.

Modelling and Simulation
AMCG

People involved J.A. Kaandorp, J. Cui.

The bioinformatics group of Amsterdam Genomics Center (AmGC) is a collaboration between researchers from SILS-UVA, IvI-UvA and Academic Medical Center Amsterdam working on bioinformatics. The bioinformatics group is coordinated by Antoine H.C. van Kampen (Academic Medical Center Amsterdam) and Jaap Kaandorp (Section Computational Science, UvA). One of the major current research themes within the AmGC is the analysis and modelling of biological networks. The bioinformatics of biological networks involves a broad range of research and approaches. This research includes topics like identification of common regulatory elements for genes in a pathway, the modelling and simulation of pathways, the reconstruction of pathways from experimental data, the visualization of pathways, and the representation of pathways in graphs and databases. To accelerate our understanding of the (dynamics of) biological networks, it is imperative that these efforts are combined. Subsequently they have to be applied to real biological problems.

Recently three projects related to modelling and simulation of biological networks have been started up within the Section Computational Science (Mesoscale simulation paradigms for biological systems, Simulation of developmental regulatory networks, Mathematics and Computation for the System Biology of Cells). In March 2004 the bioinformatics group organised an international symposium on networks in bioinformatics.

Mesoscale simulation paradigms for biological systems

People involved J.A. Kaandorp, J. Cui.

Project funded by the Applied Mathematics programme of the Dutch Science foundation, duration 2003 - 2007, total value Eur 329000

In this proposed project we want to develop and compare computational models of parts of the living cell that can calculate in detail system properties from experimentally obtained molecular and physical-chemical data. Such a model is as close as possible to the biological experiments and therefore can be used not only for understanding the principles of function but also to steer further biologicalexperiments.

Simulation of developmental regulatory networks

People involved J.A. Kaandorp, Y. Fomekong Nanfack, B. Leskes

Funded by Computational Life Sciences programme of the Dutch Science foundation, duration 2003 - 2007, total value Eur 328000

In this project we will develop a model for simulating regulatory networks that are capable of quantitatively reproducing spatial and temporal expression patterns in developmental processes. The model is a generalization of the standard connectionist model used for modelling genetic interactions. The model will be coupled with a biomechanical model of cell aggregates and used to study the formation of spatial and temporal expression patterns of gene products during development in cellular systems (sponges and scleractinian corals).

Mathematics and Computation for the System Biology of Cells (Cell.Math.)

People involved J.A. Kaandorp, J. Vidal Rodriguez

Project funded by Computational Life Sciences programme of the Dutch Science foundation, duration 2003 - 2007, total value Eur 487000

The aim of the project is to develop, implement, and validate mathematical and computational techniques for the systems biology of the cell. Biologists and mathematicians together will formulate realistic mathematical models of metabolic and regulatory networks including intrinsic spatial non-homogeneity. Depending on the cellular phenomenon considered, models and methods of appropriate temporal and spatial scales will be developed and can then be applied: models in the form of ordinary differential equations and methods for system reduction; multi-adaptive computational methods for partial differential equations (PDEs) for moderate spatial and temporal variability within a cell or an organelle; particle models describing the interaction of individual molecules and computational methods for the evaluation of the dynamic behavior; and methods for integration of these different approaches into a single simulation.

Modelling and Ananlysis of Growth and Form of Sponges and Scleratinian Corals

People: J.A. Kaandorp, R.M.H Merks

In the bio-informatics field our research in the area of marine biology also continues. Understanding external deciding factors in growth and morphology of reef corals is essential to elucidate the role of corals in marine ecosystems, and to explain their susceptibility to pollution and global climate change. A model has been developed for simulating the growth and form of a branching coral and simulated morphologies have been compared to three-dimensional images of the coral species Madracis mirabilis. With this model it is possible to generate morphologies that are virtually indistinguishable from the three-dimensional images of the actual colonies. Such models can be used to test hypotheses on mass transfer through boundary layers and the impact of the physical environment on coral growth.
BMI

People: P.M.A. Sloot

In the BMI project, we work on the modelling of HIV infections and drug therapy of HIV-infected patients in collaboration with C. Boucher of the University of Utrecht and A.V. Boukhanovsky and A.B. Degtyarev of the IHPCIS in St. Petersburg. In 2003 significant results describing the Stochastic Modeling of Temporal Variability of HIV-1 Population were published.
Nato SfP

People: A.G. Hoekstra, M Yurkin, K. Gilev, K. Semianov

In the context of the Science for Peace program of Nato we collaborate with institutes in Minsk (Belarus) and Novosibirsk (Russian Federation) on new sensitive methods for cytological analysis of heamatological samples. We concentrate on the computational science aspects and HPC simulations in the field of computational electromagnetics of the project.

LBM

A.G. Hoekstra, L. Abrahamyan

The work on mesoscopic modeling and simulation concentrated on adapting the LBM method, and specifically the L-BGK method, for unsteady flow, and to apply this to modeling flow of blood during a full heart beat in the lower abdominal aorta. We also initiated research to adapting the models to take advantage of state of the art (parallel) numerical algorithms (such as multi grids). Moreover we studied fluid-structure interaction in LBM, and have demonstrated that for 2D time-harmonic flow coupled to a simple elastic wall a theoretical expression for a dispersion relation is recovered, thus demonstrating the correctness of our model.

PECVD

Involved: V.K. Krzhizhanorskaya, P.M.A. Sloot

In this project a grid-based PSE to study the plasma enhanced chemical vapour deposition of thin films is being developed. Internally funded project in close collaboration with the IHPCIS in St. Petersburg. The research was conducted with financial support from the Dutch National Science Foundation NWO and the Russian Foundation for Basic Research under grants number 047.016.007 and 047.016.018, and with partial support from the CrossGrid EU project IST-2001-32243.
Cluster of Grapes

People involved: S. Portegies Zwart, A. Gualandris, M.S. Sipior, E. Gaburov, B. Bastijns, P.M.A. Sloot, G.D. van Albada + collaboration with the Pannekoek Astronomical Institute

One of the other research areas in which the SCS group is active, is computational astronomy. "A Cluster of Grapes", a joint NWO proposal with the astronomical institute "Anton Pannekoek" to NWO was honoured in 2001. The research covers a wide range of subjects on the boundary of Astronomy and Computational Science. As an example, Gualandris, Tirado-Ramos and Portegies Zwart studied the performance of a parallel astrophysical N-body solver on pan-European computational grids. It was demonstrated that especially for large problems grids constitute a suitable computational environment. Sipior is working to merge the “Kira integrator” with the GADGET tree code developed at MPI Garching. The end result will allow us to make use of the unique advantages of each numerical approach – the high precision of Kira for studying the dynamics of a large stellar cluster, and the computational speed of a tree code, used to simulate the galaxy in which the cluster is embedded. The first stage of this effort is now completed and undergoing testing.
Problem Solving Environments and Visualization
CrossGrid

People: A.G. Hoekstra, E.V. Zudilova, A. Tirado Ramos, R. Shulakov, D. Shamonin, P.M.A. Sloot, G.D. van Albada, M. Scarpa

In the CrossGrid project, we participate in a large European collaboration of 21 partners in the development of interactive applications in a Grid environment. These environments are characterized by the participation of multiple, geographically distributed organizations, sharing computational resources and data. This gives an improved access to these resources and data, at a cost in network and scheduling delays. In CrossGrid, we provide an application aiming to support vascular surgeons in pre-operative planning. 2004 was a consolidation of a consortium-wide integration of grid support tools and applications. The result is a fully integrated virtual vascular surgery on the grid, which will be demonstrated live during the European Grid Conference 2005 in Amsterdam.

Fig.1

CrossGrid

Token2000

People: A.G. Hoekstra, P.M.A. Sloot, E.V. Zudilova, M. Scarpa, L. Abrahamyan

Token2000 is a nationally funded project (NWO), where we collaborate closely with the Universities of Leiden and Twente on the development of an interactive medical application, somewhat similar to the work in CrossGrid. This application is intended for training of surgeons. In collaboration with Leiden University Medical Centre we have created Hemosolve, a problem solving environment for image based computational Heamodynamics. Hemosolve includes our L-BGK solver, but also a FEM Navier-Stokes solver. Moreover, it contains a 3D editing tool and powerfull visualization modules.

ViSE-VL

People: J.A. Kaandorp, R.G. Belleman, P.M.A. Sloot

In the ViSE-VL project we have developed a Visualisation and Simulation Environment, specifically aimed at medical applications. The results of the ViSE project are being used in the CrossGrid and Token2000 projects.

ViSE-VL was funded through the ICES/KIS 2 Min EZ project

High Performance Simulation on the GRID

Dutch-Russian project: NWO-RFBS-047.016.007

People: P.M.A. Sloot, A.V. Bogdanov.

In this project we study the use of Grid systems for High Performance Simulations. The project is a collaboration between Amsterdam, St. Petersburg, Moscow and Novosibirsk.

Dynamite

People: K.A. Iskra, T. Gubala, B. Ó Nualláin, D.A. Kaarsemaker, G.D. van Albada, P.M.A. Sloot.

The Dynamite project (now internally funded) is the continuation of work on the dynamic scheduling and migration of tasks in parallel programs started in the ESPRIT project Dynamite. In 2004 work continued on the single task checkpointer (ironing out some smaller remaining defects) and on developing support for MPI and grid environments.
PDES

Involvement: K.A. Iskra, G.D. van Albada, P.M.A. Sloot.

This research on parallel discrete event simulation is a continuation of the work by Overeinder on the behaviour of optimistic PDES, extended to grid-like environments. In 2004 further important results regarding the behaviour of PDES in a wide-area distributed environment were obtained. Important strategies like dedicated routers, message aggregation and lazy cancellation were studied and different methods of “global virtual time” evaluations were evaluated.
2005 and beyond
The long-standing problem of information processing in evolving Cellular Automata will be further investigated, using elements from Information Theory and Complexity theory. Our approach will be based on the notion of "local entropy" as formulated in the concept of Fischer-Information. The research will be supplemented by simulations of specific instances of Cellular Automata and will form the theoretical embedding of our CA-based modelling and simulations of biomedical phenomena.
We are currently increasing our emphasis on (Interactive) Problem Solving Environments in distributed systems like the GRID. Based upon our previous experience using HLA and the Dynamite system, our focus will be on middleware, remote access and data integration, using the currently emerging Open Grid Services Architecture (OGSA) and component technology for scientific computing (the Common Component Architecture - CCA). Participate in the CrossGrid project, a large consortium of 22 laboratories from all over Europe, with the goal to develop a software architecture for a Grid based interactive problem solving environment, applied to a number of challenging CA-based applications from Biology, Medicine and flood-crisis management. The SCS group has the final responsibility for issues related to scientific computing on the grid and the applications within the Crossgrid project. We also strongly participate in the European HealthGrid initiative.
As to the applications, we will further strengthen our attention in the Biomedical field. The main topics will be interactive simulation and visualization on the grid. A prototypical application will be vascular reconstruction. This application is part of the CrossGrid project and of the NWO- TOKEN2000 project, and is carried out in collaboration with the Leiden University Medical Centre (for the medical datasets) and the University of Twente (with cognitive psychologists, working on man-machine interaction and learning in VR). Another domain is hierarchical modelling and data integration through CA's. For example, in collaboration with the University of Utrecht Medical Center and the AMC (Amsterdam) we work in the field of HIV infection modelling using CA, and a rule-based medical expert system for HIV medication. As reported, three new NWO-funded projects in the area of Computational Life sciences have been obtained in 2003; these will play an important role in shaping our research in the coming years.
We have intensified our international collaborations. Within the framework of a formal agreement between the Universiteit van Amsterdam and the Saint Petersburg State University, Russian Federation, a joint laboratory for computational science was established in Saint Petersburg. This lab works on problem solving environments and applications. We formally collaborate with CYFRONET, Cracow, Poland, through joint Ph.D. students in the field of monitoring and scheduling for scientific computing on the Grid. In the framework of a recently started NATO Science for Peace program we collaborate with institutes of the Siberian Branch of the Russian Academy of Science in the field of biomedical diagnostic systems. Finally, strong research cooperation is established with a number of USA based universities (Stanford University, Mississippi State University).
Relation to other programmes

We collaborate with the Parallel Systems Architecture and Programming Methods and Environments groups in the Computing, System architecture, and Programming Laboratory (CSP-Lab).

Through the Virtual Lab project we strongly collaborate with the Parallel Systems Architecture group. The theoretical work on evolving CA is done in collaboration with the Algorithms and Complexity Theory group of the ILLC.