Sunday, January 22, 2017

Report on the rest of SOFSEM 2017 (guest post by Ignacio Fábregas)

I continue my notes on this year's SOFSEM conference where I left them. On Wednesday, I attended the tutorial session by Andrew Butterfield on Unifying Theories of Programming (UTP), a paradigm introduced by C.A.R. Hoare. Broadly speaking, UTP combines denotational, operational and algebraic semantics in a unified framework for both the formal specification and implementation. It is almost as there was no distinction between syntax and semantics. Andrew Butterfield also showed us  his tool UTP2, programmed in Haskell, and how the library for UTP in the Isabelle theorem prover works. 

One important part in any conference is to be able to discuss with other colleagues and experts about interesting papers. On Wednesday I missed the keynote talk by Jakko Hollmen about Predictive Analytics because I was talking about some papers with colleagues. (I found that to be a valid excuse to miss a talk in a conference!) Also, on Wednesday we had the social programme: an excursion with a guide around the city centre of Limerick and the conference dinner in the Bunratty Castle. Inside the castle we not only enjoyed a typical Irish dinner but also an entertainment program. It was a dinner with a medieval flavour because of the place but also because of the show we saw and the way we ate. We had a broth, pork ribs and chicken... But without forks! So we ate with our hands while listening to some typical Irish songs performed by the very talented people of the Bunratty Castle.

That's not all that happened on Wednesday. The organization had a surprise for us: since the lunch break was two hours long, during the second hour we had a talk about music and computers. Inside the Department of Computer Science in Limerick there is the Digital Media & Arts Research Centre, and the organization took advantage of this fact in order to propose them to give 3 talks: one on Wednesday and two on Thursday. The first talk was done by Mikael Fernström (http://www.idc.ul.ie/mikael/), who explained us his work. He uses computers in order to produce new sounds (frequencies that a physical instrument wouldn't be able to make) and algorithms in order to create music for "things". As an example, he created a piece that took some variables from meteorological data so we could listen to how the weather in Ireland sounds.

On Thursday, in this special sessions, Giuseppe Torre (http://www.muresearchlab.com/) showed us his work, where he uses computers to represent art. For example, his work 'AI PRISON' is a program in C++ that outputs '0' unless the technological singularity in Kurzweil's sense (or 'Terminator' sense) appears. It is exposed in the BLITZ Contemporary Art Gallery in Malta. Finally, Kerry l. Hagan (http://www.kerrylhagan.net/) played some of his musical works. Again, the use of computers is the cornerstone. She uses a program that allows her to program and compose the music she is going to play. This three artistic sessions where a really nice touch from the organization.

Back to the science, on Thursday we started with a keynote talk by Óscar Pastor López about Model-driven Development. The idea is that developers should be freed from programming concerns and be able to concentrate on guaranteeing that the final software product corresponds to what the company asked for. There are two basic notions in this field: capability and the Model Driven Architecture. The first one is especially used at the earliest steps of a software process in order to characterize the relevant modelling components to be specified. Whereas the Model Driven Architecture is used in the definition of the desired software process in order to structure the development from requirements to code.

Afterwards I attended the tutorial by Cristina Seceleanu about verification and test-case generation for Automotive Systems. Nowadays a car is developed with more electric components than hydraulic ones. For example, the connection between the steering wheel and the steering control is done by wires and software. This makes mandatory the use of testing in order to verify the correct behaviour of a car. In this field of research tools as UPPAAL are of great use. After the lunch we had the second keynote talk of the day. Paul Spirakis explained us about "programmable matter", which at the beginning seemed a field closest to sci-fi than actual computer science. So, let me explain what I understood, nowadays digital information (as internet) has spread along the world. That information is simply data, which can be manipulated. This vision is focused in our ability to manipulate data, and thus extended to manipulate matter via information-theoretic and computing mechanisms, incorporating information to the physical world. This is inspired in the biological world where living organisms are built by cells and those cells are built by information (DNA). Paul Spirakis explained us how Network Constructions is a promising technique for modelling this idea of "programmable matter". Broadly speaking, Network Constructions is simply a collection of finite-automata that interact randomly like molecules according to the rules of a common protocol.

The last session on Thursday for me was the tutorial session by Louis-Marie Traonouez by Plasma Lab, a statistical model checker (SMC). A SMC and uses less resources than a model checker at the cost of losing certainty. SMCs answers with a confidence level. Plasma Lab is SMC a library written in Java that can be used in tools as UPPAAL. The properties can be described in a form of bounded linear temporal logic (like LTL but with bounds in the length of the paths).

The last keynote talk of SOFSEM 2017 was on Friday and by Igor Walukiewicz. The topic of the talk was automatic verification of recursive programs with thread creation. Although this situation appears in programming languages such as Java, Scala or Erlang it doesn't behave well in automatic verification. For example, reachability is not decidable even for pushdown systems where two threads are communicating over a 2-bit shared variable. Igor Walukiewicz showed us some decidable setting by relaxing the semantics of thread creation operation. The obtaining result might not be very efficient but at least is computable.

Finally, I want to end these notes on SOFSEM by acknowledging the excellent work of the local organization at Limerick. It's true that SOFSEM 2017 didn't start very well in June, but all things considered everything went smooth and fine in Limerick. Also, the organization always tried to add something new to the conference (as those special talks about art and computer science), which is always nice for the participants.

Next year SOFSEM will be held in Austria and my recommendation is to check it out!

No comments: