The Técnico PhD student and CERENA researcher Helga Jordão was one of the 25 brightest minds selected to join NASA’s Frontier Development Lab (FDL). The researchers were selected among 257 candidates and will have to create a digital model that allows to predict coastal evolution and dynamics, through machine learning, in a faster way and with less computational requirements than the current ones.
FDL is a public-private program organized by NASA, in partnership with the SETI Institute, with support from Google Cloud, IBM, Intel, Microsoft and NVIDIA.
Helga Jordão is the only Portuguese researcher and she is “incredibly happy to have been selected”. “I must have read the email saying that I had been selected 3 times to make sure I was reading it correctly”, recalls Helga Jordão.
The program started 2 weeks ago and will extend for another 6 weeks. According to the Técnico student, “it has been intense and challenging so far”. “During the first week we had to define the possible approaches to the challenge, from a simpler and therefore safer approach, to a more complex and therefore difficult approach, in order to implement solutions. At the same time, we had training in various Cloud platforms”, she says.
The teams are in contact on a daily basis, but the experience has no delimited borders, and there is also a connection with the members of the other teams and partners. “We exchange ideas and advice and we feel that we are part of a big community of experts in different areas, in which everyone is interested and focused on the different challenges”, says the CERENA researcher. “It’s been wonderful to work with people so full of ideas and knowledge”.
In emergency situations such as those caused by a sudden rise in sea level, it is essential to quickly access forecast maps to enable the taking of measures to minimize the impacts caused by this type of phenomenon. Current models, although accurate, require a lot of computational power and a lot of time to simulate. A digital model of coastal dynamics will significantly reduce simulation time and will allow, in emergency situations, to better manage the situation. Another advantage will be the reduction of the high computation capacity required, which means that the simulation of these models is currently inaccessible to some countries, countries without high computation infrastructure, which are frequently victimized by these natural phenomena, for example the Philippines. “Our goal is to create a digital model through machine learning to replace the existing models, but much faster, in order to allow its use as an operational tool, and with less computational requirements, so that its use is not restricted to certain countries”, explains the the PhD student.
Helga Jordão’s contributions are directly related to the understanding, in physical terms, of meteorological phenomena and how they interact and affect coastal zones. The CERENA researcher brings the experience gained from her doctoral thesis, which focuses on the modelling of complex mineral bodies through the application of deep neural networks.
During these weeks, and with the support of mentors from NASA, Institute for Simulation Intelligence, MIT (+ MIT Portugal), Planet’s Analytics Applications, USGS and Berkeley Lab, the young researchers – of various nationalities – will try to develop this innovative and pivotal tool, and an artificial neural network. “We intend to use Physics Informed Neural Networks, neural networks consistent with the laws of physics, characterized by non-linear partial differential equations”, explains the researcher.
All the knowledge developed during this program will be made available to the community, free of charge and in open source. “In the specific case of our challenge, one of the objectives is to create a computationally light digital model so that the use of the model is not limited to some countries. And this would not make sense if it were not open source”, stresses the researcher.
