Scientists from several European and US institutions authored an article published in Nature, last May, which indicates their position in favour of using machine learning techniques to High Energy Density Physics. Marta Fajardo, professor at IST Department of Physics (DF) and researcher at the Institute for Plasmas and Nuclear Fusion (IPFN), is one of the authors of the article “The data-driven future of high-energy-density physics”.
“This is a ‘Nature Perspectives’ article where a large majority of the community shares their perspective and reaches a broad consensus on the use of new machine learning techniques in the interpretation and design of extreme plasma experiments”, says professor Marta Fajardo.
“High-energy-density physics is the field of physics concerned with studying matter at extremely high temperatures and densities. Such conditions produce highly nonlinear plasma, in which several phenomena that can normally be treated independently of one another become strongly coupled. The study of these plasmas is important for our understanding of astrophysics, nuclear fusion and fundamental physics—however, the nonlinearities and strong couplings present in these extreme physical systems makes them very difficult to understand theoretically or to optimize experimentally”, the article’s abstract says.
According to the Técnico professor “artificial intelligence can be applied in several ways: on the one hand, it can help to optimise experiments, using machine learning to improve the generation of X-ray sources, accelerated particles, etc., which are necessary for the most demanding diagnoses”. “On the other hand, it can efficiently use all the data generated by the experiments, which nowadays have high repetition rates, and therefore generate floods of data, to establish correlations and verify with a much higher degree of certainty the theoretical magnitudes we aim to determine”, stresses the researcher. “Finally, deep learning techniques can help to generate simplified models, which replace what we can call “hero simulations”, which are computational models in which we try to introduce as much physics as possible but, for this reason, they are extremely heavy from a computational perspective and do not allow for a very large parameter sweep”, adds the professor.
A team of researchers from Imperial College London, using the Central Laser Facility (CLF) of the Science and Technology Facilities Council, showed that Artificial intelligence improves control of powerful plasma accelerators and was able to optimise the accelerator much more quickly than a human operator. The IPFN researcher explains “while a human operator tends to make a linear parameter scan, an algorithm can optimise many parameters at once until the optimal parameter configuration is reached. But the truth is that a highly skilled operator ends up getting very good results”.
The scientists involved in this work estimate that the use of AI will contribute to go beyond our knowledge and understanding of extreme plasma physics.
The use of these AI tools could be used in VOXEL project and, therefore, professor Marta Fajardo, VOXEL coordinator at Técnico, has already submitted a project that aims to automate the entire line of interaction, from laser to diagnostics. “The idea is to spend as little time as possible optimizing X-ray or electron sources, and devote ourselves to collecting new data with new imaging techniques we develop here. I am keeping my fingers crossed for it”.
