ADASS 2011 - the other days
So, as expected, a daily post about ADASS would end up being too much. The daily program was too long and then comes the social part of it ;-)After the tutorials there was the opening reception in the Observatory, in the beautiful Cassini Room. I really have to go there again with a real camera hehehehe
Summarizing what I found interesting in the talks.
There was a lot about GPU processing. GPUs are able to do floating point operations with much better than CPUs, and they have lots of processors inside, so with parallelization one can really improve processing (on what can be distributed). BTW, don't ask the GPU to write a text, it's not able to ;-)
Nice processing example using GPU would be the GPU version of the FATBOY (University of Florida) pipeline.
Another point was Cloud computation. For several cases it is good and applicable. Instead of buying high power machinery, use Amazon, for example. Might be faster (since you don't have to draw specs, approve budget, collect invoices, buy the H/W, install all the S/W and fine tune it ...
VO world goes on, despite ESO going from active to passive member of IVOA. Lot's of interesting things going on. At ESO we need to find a way to extend VIRGO's life ... let's see what can be done.
Lot's of smaller communications on on-going projects. This would have been the perfect place to present the OV_WAV. Well, it's not all lost, we still can finish the paper and present it there!
Synergy between different areas was a lesson to take home. Eric Feigelson told us how he made statistical analysis of quasars using the statistical methods developed for survival rate analysis in biomedicine.
Sometimes all it takes is mention the problem to someone.
Some years ago I wanted to do automated search of compact groups in SDSS. The idea of calculating distances from each galaxy to all the other ~140 million galaxies in the catalogue was out of the question. There had to be a better way (I know there are several, I just needed to know which). There was. I mentioned the idea to my office mate at ESO, Joerg Dietrich, and he mentioned KD-Trees and that he was building a library on that a while ago. We just had to include spherical coordinates in it and voila, there it was, the solution.
I was not aware of binary or any other kind of tree, but mentioning got me the solution. The project didn't go forward for other reasons, but we had the tool.
That's why coffee is so important hehehehehe The solution might sit in the same office as you ... you just need to find it out!
posted by Cris Da Rocha @ 3:07 PM
0 comments
