More About This Event
What are the main applications of Seismic Inversion in your research centre?
A.V. – In our team, two research trends have been living together for 10/20 years:
- Seismic tomography;
- Seismic modeling (the major international expert is Mr. “Jose Carcione”).
Considering these two branches of geophysics, which we have been working for about a year and a half to build our own approach to full waveform inversion. We do not want to replicate what already exists but look for new ways to reduce computational cost.
In particular, we have been thinking since the beginning to develop a parallel code. We have proposed some projects to keep multi-year funding of the ERC (European research council) and other public or private investors.
In your opinion, what are the future goals of Seismic inversion?
E.D. – In my opinion, researchers have recently been developing parallel algorithms of FWI, using tools such as High-Performance-Computing to implement Machine Learning techniques to analyze Big data (i.e. genetic algorithms).
A.V. – in the future, we will not have to limit ourselves to estimating only the P speeds but we will have to work more on the elastic front and extend the analysis also to the inelastic field. We could also get good results on the anisotropy front, but for the moment, I remain a bit skeptical. With a view to reducing calculation times, as Eleonora said, I believe that artificial intelligence is a very promising new frontier.
As an old physicist, I love this technique very much. I hope that the use of artificial intelligence is a way to improve the physical / analytical knowledge of the problem, constituting a more robust approach to represent the nature that surrounds us.
You both mentioned parallel computing, artificial intelligence, machine learning; what has been the impact of these technologies on today’s geophysics?
E.D. – nowadays the massive use of artificial intelligence has almost become a fashion. As Aldo said, techniques such as Neural Networks should not be used immediately. First, it is necessary to try to simplify the problem from an analytical point of view, defining sensible optimization functions. Already from the beginning, you can do a lot without immediately resorting to super calculators.
A.V. – I absolutely agree. I would like to add a historical note … I remember that in the 90s, expert systems were very fashionable, it seemed that these would be the future of Geophysics. In reality, after 10 years of talking about it, the revolution has not been seen, now there has been 20 years in which no one has spoken about it anymore.