European computer scientists seek new framework for computation

Tag : computer
By European Science Foundation, [RxPG] There have been several revolutions during the 60 yearhistory of electronic computation, such as high level programminglanguages and client/server separation, but one key challenge hasyet to be fully resolved. This is to break down large complexprocesses into small more manageable components that can then bereused in different applications. There are many possible ways ofdoing this, but none of them cope well with all processes, with themajor problem lying in the dependant links, or correlations,between components that cannot be broken down, the threads thatinterconnect whole computer processes or programs. Thesecorrelations are common to all processes in which computation isinvolved, including biological systems and the emerging field ofquantum computing, as well as conventional programming.

European computer scientists believe the time is right now for acoordinated effort to solve the correlation problem and a group ofthem recently held a workshop organised by the European ScienceFoundation (ESF) to establish a framework for further research. Theworkshop was an astounding success, firstly in identifying thatcorrelations in computer science represented an important problemcommon to the whole field of programming and software developmentnow highly relevant to all industries and everybody's lives. Itwas, as was noted by the workshop's convenor Ellie D'Hondt, aspecialist in quantum computing research at Vrije Universiteit inBrussels, an important forum for accumulating the requiredexpertise to take the field forward.

We are now at a stage were all participants understand why we needa correlation paradigm, that there is a commonality between thefields included, and we converged on a definition and basicprinciples, said D'Hondt. People are now ready to do research onthe problem, and this is what we should get together on in anotheryear or so.

Now is a good time to tackle the correlation problem. The evolutionof general purpose computing has reached a point where thecorrelation problem can stand in the way of progress. The explosionof the Internet has been associated with rapid growth in softwarecomponents designed to be reused to avoid the cost of duplicatedprogramming effort.

The workshop discussed progress in the relatively new field ofaspect-oriented software development (AOSD), which is bringing newtechniques for isolating the correlations cutting across softwarecomponents. The techniques of AOSD make it possible to modularisethose aspects of a system or process that cut across differentcomponents. In this way the cross cutting aspects themselves can bebroken down into reusable components or objects. This in turnenables a whole process to be broken down more completely intocomponents that also embrace the cross cutting aspects.

Research into correlation is also timely because expertise isemerging independently in three different fields, quantumcomputing, bio computing, and AOSD, the latter being mostapplicable to general purpose computing. As D'Hondt noted,cooperation between specialists in these fields is needed to avoidduplication of effort, but more particularly because it willstimulate and drive forward the whole study of correlations. Onthis front the ESF workshop was highly successful, because itbought together representatives from each of the three fields insmall groups. It was amazing to have these groups of peopleactually communicate, said D'Hondt. We split up into small groupswhere there would be one aspect, one quantum and one bio person,people not usually knowing each other beforehand, and this worked!People came up with small presentations after only one day of talksgetting introduced to the whole body of work.

A common thread emerged from these mini-workshops, which was thefact that correlations appear when progressing from the high levelglobal description of a problem to the lower level localcomponents.

Correlations capture the interaction between the parts, saidD'Hondt. In other words the devil is in the detail. This is as truein biological systems as say a web based search engine. In thehuman brain for example it is possible to define how long termmemories are formed, but this does not tell us how an individualneuron might be phase locked with another at a local level, so thatthe two depend on each other. Similarly in computation, a highlevel view does not describe the particular order in which lowerlevel components need to be executed on the basis of thecorrelations or links between them. For example two sub-programsmight share a common variable, which decides when they have to beexecuted within a larger task or application.

The ESF workshop also established a common theme, which was thatcorrelations can be a good thing, rather than a hindrance tocomputation, as has been shown in quantum computing. Correlationsare often seen as a burden, a nuisance, something making theproblem hard to solve, said D'Hondt. But my experience in quantumcomputing tells me it is something that can also steer computationsor even make them possible.