From ICCSA, Perugia, Italy

Computational Science and Its Applications (ICCSA) is a research society that meets regularly every year ever since 2001. The last seven meetings were held in Kuala Lumpur, Malaysia (2007), Glasgow, UK (2006), Singapore (2005), Assisi, Italy (2004), Montreal, Canada (2003), and (as ICCS) Amsterdam, The Netherlands (2002) and San Francisco, USA (2001). The 2008 conference was held at the (Campus) University of Perugia at various locations (Department of Mathematics and Chemistry mainly). The lecture rooms had no air-conditioning (and it was really really hot), so I wondered why did they had to put the conference in such a warm month. The other frustrating issue was orientation during the first two days (almost all events --- workshops, coffee breaks, plenary lectures, registration etc. --- were at different locations); first day, beginning was more than 40 minutes delayed. The conference was split into numerous sessions, ranging from urban analysis to forrensincs and pure IT applications. The titles of papers seen at the conference was extremely diverse --- they ranged from gaming industry ("Web usability on the Nintendo Wii Platform"), to spam filtering ("Identifying Spam Web Pages Based on Content Similarity") and forensics ("Crime Scene Interpretation Through an Augmented Reality Environment". According to one of the organizers (Osvaldo), this meeting was almost exclusively dedicated to geography/GIS applications as there were more than 45 presentations that were addressing computations with spatial data. In principle, I only followed the two geo-sessions: "Geographical Analysis, Urban Modeling, Spatial statistics" and the one that I co-organized with a group of colleagues from the Netherlands "Computational Geomatics".

There were six keynotes. Dennis Pumain (Université Paris) presented trends and developments in the area of human geography and urban modelling (mainly at European scales). Urban geographers are actively working on developing urban growth simulation models --- I did not know that one can simulate growth of cities and city networks. I managed to note only some of the many interesting URLs: [http://s4.parisgeo.cnrs.fr], [www.spatial-modelling.info], [www.cybergeo.eu], [http://libercarto.prd.fr]. Pumain just published a book on the topic entitled "Hierarchy in natural and social sciences". Kenneth Tan (Optimanumerics Ltd.) discussed importance of supercomputing and listed a number of actual issues (mainly IT-connected). Tan: "the reason for supercomputing is to get results sooner, with more accuracy... and retire sooner". In the context of globalization, "the idea of supercomputing is to significantly reduce cyber-distances signals have to cross (at a speed of light)". However the market for computers looks diverse and dynamic, there are still only for vendors of chips: Intel, AMD, IBM and Sun. Campos Plasencia (Istituto de Fisica de Cantabria) further gave an overview of the GRID projects in Europe and discussed many interesting topics. Plasencia: "E-science can be defined as collection of scientific activities that are carried out by using resources distributed over the internet --- so everybody today is doing E-science? Not true --- E-science can be only connected with applications where utilization of internet adds value --- there has to be a necessity to employ internet". The question remains when does a project becomes an E-science project? Today, internet applications and mobile technologies are intermixing so that it is becoming increasingly interesting "how to enchance/support collaboration between multidisciplinary projects in post-internet area?". Plasencia defines GRID (computing) as: "a set of IT resources that can be used in a combined way to solve excessive scientific computations". There is a difference between "supercomputing" and "GRID computing" in a sense that GRID computing assumes larger networks of computers (more CPU hours; 'capacity computing') and supercomputing is all about high computing speeds (GFlops/s) --- even on a single computer. GRID networks can be best compared with the electricity networks over continents. Plasencia cited Ian Foster: "the true revolution was not the discovery of electricity, but the power to distribute it". In that sense, distribution of electricity is much better solves: "anybody can plug-in, obtain electricity and then pay for exact electricity usage; with GRID computing we are still not able to do the same". Basis of GRID computing is the existence of so called "middleware"; in Europe the most known is gLite --- this is a sort of a 'CPU resource broker' that looks for resources on a GRID network and then distributes the jobs. A problem of gLite is that is not designed to feed into more than one infrastructure. This asks for more standardization among the GRID computing communities (e.g. www.ogf.org). Plasencia: "Universal GRID middleware is a real challenge! On one thing all GRID developers agree --- the strategy needs to always be: hide complexity from the final user to enhance the added value of using GRID computing". In Europe, there are about 20 GRID computing projects, most of them are actually quite focused: biomed/pharmacy, CERN, bioinformatics, space research etc. For example, a GRID computing project that is currently being implemented is developed to support the "Planck Mission" --- the new satellite for background radiation mapping and testing of the Planck sky model. Plasencia: "there is no single GLOBAL GRID! You always need to be a member of some research team". In any case, GRID computing is a future: from June 2007 to July 2008, EGEE registered 150 millions of CPU hours, which is several times more than few years ago. Recently, there has been a big push (see also RI) to combine all GRID computing projects in Europe into a large global network called EGEE (for more info see www.eu-egee.org or www.interactive-grid.eu; in USA, a web-portal to enhance GRID computing called TeraGRID exist). Plasencia further showed some actual examples where GRID computing is used in an operational way. I noted two: (1) mapping of anomalies of chlorophyll on the sea surface in the Mediterranean sea [http://www.bo.ingv.it/mfstep/] --- this is a combination of sophisticated sensors called "GLIDERS", small automated submarines that dive and go to the surface and emit measurements of various ecological parameters; these measurements are then used to generate maps and as warning systems. (2) monitoring toxic algae blooms in inland waters (lakes); again, a network of automated sensors is used as a warning system; for more info see http://www.dorii.eu (Deploying and Operating Remote Instruments).

Vipin Kumar (presented by Osvaldo as "the farther of many of us") focused on computing issue in data mining applications. Kumar: "Pentaflop computers have arrived!" (e.g. Jaguar Cray XT4/XT3 Oak Ridge NC). Thomas Watson said in 1943: "I think that there is world market for maybe 5 computers". Kumar "We need supercomputing because the sensors are producing increasing quantities of digital data. The sensor technology is putting a big pressure on computers". The computing is increasingly data-driven --- such data can simply not be processed using conventional IT. Some examples of computational kernels and data mining are: (1) distance computation; (2) statistical parametrization, (3) structure detection. Kumar further presented many DM mapping applications. (1) change detection based on the MODIS EVI images (SF Bay area). The accuracy of detecting the land cover change areas was further validated using GE. In all cases the detected areas shows to respond to increasing urbanization. (2) data mining can also be used for climate monitoring and detection of anomalies in meteorological images, such as El Nino and similar. (3) another interesting application was use of the global 4 km resolution vegetation images (FPAR) for monitoring of large fires and similar excessive events. A list of recent publications of Kumar can be found at: http://www.cs.umn.edu/~kumar

I also participated at the meeting of the conference organizers and discovered many interesting things. The organizers had to actually pay Springer to publish the special issue. I could not find out how much. In my opinion, costs of the conference was relatively high considering all facility problems and a low number of key note lectures. On the other hand, it was nice to receive the conference proceedings at the beginning of the conference.

There were many interesting papers/presentations, especially coming from PhD (even graduate) students from various countries. These are some titles potentially interesting to ISARA:

A complete list of papers published in the proceedings are available at: [http://www.springerlink.com/content/r1x1l475q3r5/]

On the one-before-last day of the conference, a group of vocals has joined us and performed pieces from the most famous Italian operas ("Le nozze di Figaro", "La Boheme" etc.), followed by the traditionally long and impressive tasting of Italian food. All in all, I liked a lot that the conference had a real academic flavor and that it was held in this beautiful city (much better than some expensive hotel). I also enjoyed meeting Vipin Kumar, Mikhail Kanevski and a number of Dutch colleagues (e.g. Cor Veenman, a forensic). I still have the impression that choice of the month/location and the production/distribution of the papers could have been better. Unfortunately, scientific groups are often hard to organize and motivate for open access publishing and quality review. In the case of ICCSA I definitively did not like it that they actually pay Springer to publish the proceedings (this was not specified on the website). The next ICCSA conference will be held in 2009 in Korea.

Location: 
Perugia campusPerugia
Italy
43° 6' 59.9904" N, 12° 23' 3.5808" E