(Machine) Learning How to Sea: Meet Raska
An excited Raska on a beautiful English day. 
Manchester, UK – the birds are chirping, and the sun is unusually cooperative in England this morning. I get up and commence my daily routine, which obviously includes the mandatory tub of hot coffee. It is the start of PRACE Summer of HPC, and I am ready to enjoy my high-performance summer!
A little bit of background about myself; my name is Raska, I am 21 years old and I come from sunny Jakarta, Indonesia. Over the past few years, I have been studying at the University of Manchester in the UK where I recently obtained an undergraduate degree in aerospace engineering.
I was extremely happy when I received my invite for Summer of HPC – not only am I given the opportunity to work on supercomputers, but also in such a diverse setting, with students and host institutes spread all across Europe. Our international team is made up of 5 individuals with an interdisciplinary set of specialisations. Working alongside me are; Mario who is a physicist from Germany, Veerle and Silvia who are both marine geologists at the UK National Oceanography Centre and the Italian Institute of Oceanography and Applied Geophysics (OGS) respectively, and Massimiliano who is our HPC specialist based at CINECA in Bologna, Italy. Of course, the current pandemic has made it difficult for travel, even within Europe. Therefore, I will be working from my room in Manchester.
Automated classification for mapping submarine structures by artificial intelligence strategies
This summer, I am joining CINECA as part of the Summer of HPC programme to work on using artificial intelligence methods to classify submarine structures of the Mediterranean Sea near the Italian coast (Project 2111 – see more details here). The idea is to develop a system that can learn from depth maps of the sea to automate the classification process of undersea reliefs. The classified maps are then used to help identify marine geohazards, which can help a range of applications such as search and rescue, habitat mapping, and more.
I became interested in this project after doing my undergrad dissertation; I developed software that detects and learns the shape of damage in structures such as cracks and vegetation to help engineers conduct inspections of buildings and tested it on a 3D laser scan of a bridge here in Manchester (you can check out a short overview here). While the sea can be very different to a building, computer vision principles can be transferable across different datasets. So, when my dissertation tutor encouraged me to apply for this programme, it was a complete no-brainer for me.
Supercomputers to the rescue
Classifying undersea structures of the Italian coast is obviously more computationally intensive than cracks on a bridge (spoiler: the former is far much larger!), so to achieve this goal we will need the appropriate resources. While I sit here in my room, our program will run on Marconi100, one of CINECA’s supercomputers in Bologna, Italy. Supercomputers are basically a cluster of computers (called ‘nodes’) that work together solve problems. Marconi100, as of July 2021, is built up of 980 nodes – imagine how powerful it is!
Before we go…
I’m going to end this post with a little question; I previously mentioned how supercomputers such as Marconi100 are powerful enough to run artificial intelligence software on big sets of data, but that is only one of the many things it can do! What kind of problems do you think supercomputers work to solve? I am interested to see what you think!
 Raska Soemantoro – Author’s own image
 Little computer and super computer – Clips Ahoy Free Images http://www.clipsahoy.com/webgraphics2/as2419.htm
 Free question stock photo – FreeImages https://www.freeimages.com/photo/question-1239547