Conclusion

The growing amount of data characterizing physical systems, whether derived from experimental tests, in situ measurements or numerical simulations, and the ongoing development and improvement of algorithms for compressing and exploiting this data, enable us to build numerical models that are increasingly predictive of the behavior of these systems in certain configurations.

The hybridization of digital simulation with machine learning is renewing scientific computing practices and pushing back some of the current limitations of modeling: more and more precision with fewer and fewer computational resources. At least at this stage in the development of machine learning techniques, it is probably not conceivable to do without equations and models, which synthesize knowledge (constructing a phenomenon, a system, etc.). However, it is interesting to realize that a new...