After a 2 week holiday, I’m finally back to work. Before letting more time pass by, I would like to share here a small summary of the e-Science conference I attended about a month and a half ago in Chicago.
I’ll start with the keynotes. There were four in the 3 days that the conference lasted. Gerhard Klimeck (slides) introduced Nanohub, a platform to publish and use separate components and tools via user-friendly interfaces, showing how they could be used for different purposes like education or research in a scalable way. It has a lot of potential (specially since they try to make things easier through simple interfaces), but I found curious how the notion of workflows doesn’t exist (or they are barely used).
Gregory Wilson (slides) raised a nice issue in e-Science: sometimes the main issue about the products developed by the scientific community is not that they have the wrong functionality, but that users don’t understand what are these products or how to use them. In order to address it, we should first prepare the users and then give them the tools.
The third speaker was Carole Goble (slides), who talked about reproducibility in e-Science and the multiple projects in which she is participating. She mentioned specially the wf4Ever project (where she collaborates with the OEG) and the Research Objects, the data artifacts that myExperiment is starting to adopt in order to preserve workflows and their provenance.
The last keynote was given by Leonard Smith (slides), and unlike the others (which were more computer science oriented), he presented from the point of view of a scientist that is looking for the appropriate tools to keep doing his research successfully. He talked about doing “science in the dark” (predictions over past observations) versus “science in the light” (analysis with empirical evaluations), and showed the example of meteorological predictions. Apparently the Royal Society wanted to drop the weather predictions in the past, but they were forced by users to have them back. Leonard highlighted the importance of never giving a 100% or 0% chance in the forecasts and ended his talk asking how could the e-Science community help this kind of research. I really recommend taking a look at the slides.
As for the panels, I attended the one about operating cities and Big Data. The work presented was very interesting, but I was a bit disappointed. I haven’t been to many panels before, and I thought a panel discussion was more a discussion between the speakers and the audience rather than presentations about the speakers’ work and a longer round of questions. This does not imply that the work was bad at all, just that I missed some debate among the invited speakers.
Regarding the sessions, most of them happened in parallel. The whole program can be seen here, so I will just post those which I enjoyed the most:
- Workflow 1: Where Khalid Belhajjame presented the work on decay analyzed by the wf4Ever people in Taverna workflows (slides). Definitely a good first step for those seeking to preserve the workflow functionality and their reprpoducibility. In this session I also talked about our empirical analysis on scientific workflows in order to find common patterns in their functionality (see slides).
- Data provenance: Beth Plale’s students (Pend Chen and You-Wei Cheah) introduced their work on temporal representation and quality of the workflow traces; and Sarah Cohen-Boulakia presented her work about workflow rewriting in order to make scalable analyses on the workflow graphs. I liked all the aforementioned presentations, as they where interesting and easy to follow. However they all shared the need on real workflow traces (they had created artifical ones for testing their approaches).
- Workflow 2: From this session I found relevant the work presented by Sonja Holl (slides), who talked about the approach they use to find automatically the appropriate parameters for running a workflow. Once again, she was interested for traces o real workflows, specifically from Taverna (since it is the system she had been dealing with).
In conclusion, I was very happy to attend to the conference (my first one if I don’t count workshops!), even if I missed the 3 day workshops from Microsoft that happened earlier in the week. I had the chance to meet new people that I had only seen through e-mail, and I talked to all the thinking heads working close to what I do.
From the sessions also became clear to me that the community is asking for a scientific workflow provenance curated benchmark for testing their different algorithms and methods. Fortunately I have seen a call for paper with this theme: https://sites.google.com/site/bigprov13/. It covers provenance in general, but in the Wf4ever project we are already planning a joint submission with more than 100 executions of different workflows from Taverna and Wings systems. Specifically, the ones from Wings are already online published as Linked Data (see some examples here). Lets see how the call works out!