It's about 25 years ago with something called 4D variational data assimilation.

这大概是25年前的一种叫做四维变分据同化的东西。

Variational inference, that math comes from statistical physics.

变分推，那来自计物。

A lot of these modern learning methods, or VAEs, which are variational autoencoders, you can really explain them very nicely in terms of an energy function that they're trying to minimize or maximize.

这些现代习方法，或者说是变分自编码器（VAEs），你可以非常巧妙地用它们试图最小化或最大化的能量函来解释它们。