A good machine learning algorithm is a powerful research accelerator. Combined with a computer simulation, it can find mathematical shortcuts in the program, helping scientists more quickly understand the effects of drugs on cells or the potential of rocket engines to carry humans to Mars and beyond.

New research is giving scientists around the world this tool. In a recent article in npj Computational MaterialsA team of researchers from Sandia National Laboratories and Brown University have presented a universal method for accelerating virtually any type of simulation.

“From the user’s perspective, there is no difference between running your simulation tool and running this accelerated simulation tool. You get exactly the same predictions. The difference is how much time it takes to get those results,” said Rémi Dingreville of Sandia.

Dingreville and his team used their accelerator to run a simulation for materials science 16 times faster than usual. And more importantly, they outlined in their article how it could be used just as easily to accelerate computer programs for climate research, the navigation of self-driving vehicles, or hardware acceleration.

“The ability to generalize our approach to different systems could lead to more efficient and sustainable technologies,” said Vivek Oommen of Brown, lead author of the paper.

Accelerator democratizes fast science

As a child, Dingreville loved to be fast. He rode his bike fast, skied fast, and ran fast. He even competed to finish his schoolwork first. Today, as a scientist, he uses machine learning to speed up his research. In a previous project, he redesigned a simulation that ran 40,000 times faster.

While a 16x speedup may seem modest in comparison, Dingreville and team stress that their latest work could have a much larger impact because it benefits virtually every field of science. It is not limited to specific types of problems like other accelerators.

“Physics, chemistry, geochemistry, weather forecasting – it really doesn’t matter,” Dingreville said.

The team sees their paper as a challenge to researchers to fundamentally rethink the way they design and use simulations.

“I am deeply fascinated by the challenges and potential of integrating traditional numerical methods with artificial intelligence to solve complex problems in materials science,” said Oommen.

Faster simulations enable new research opportunities

The Simulation Accelerator not only saves time and money on routine research, but also removes obstacles to studying phenomena that cannot normally be simulated. Try to model an event that develops slowly, such as the melting of glaciers, and your program will probably take too long to be useful.

“Currently, you have to resort to these direct numerical solvers. Even though they are accurate, they are slow,” said Dingreville.

The team hopes that this research will mark the birth of a modern, common method for scientists to quickly perform normally sluggish simulations.

“I am excited to see how our methods can be applied to other challenging problems in various fields such as energy, biotechnology and environmental sciences in the future,” said Oommen.

“I would like to see this applied in geoscience,” Dingreville added.