EN ▾
SNAPE REP RT Front page →
ConfirmedScience & maths20 Nov, 20:00 UTC

MIT method guarantees AI models of chaotic systems such as weather and turbulence stay physically bounded

Illustration from MIT of a chaotic trajectory kept within a learned boundary by an energy-dissipation constraint.
Sunbochen Tang / MIT

In PNAS, Sunbochen Tang, Navid Azizan and Themistoklis Sapsis enforce energy dissipation as a hard constraint so long simulations cannot grow without limit while still behaving chaotically; MIT says it is the first work with such formal guarantees for machine-learning models of chaos.

Read at MIT News Share

Sources

More on Science & maths
More from Snape ReportFront page →