Replies: 2 comments 3 replies
-
I'm sure it can be done. In Eureqa, it was possible to express the following The built-in function D is a derivative (2 means 2nd derivative), and sma is a smoothing averager. soi is a time-series data |
Beta Was this translation helpful? Give feedback.
-
Thanks @pukpr and sorry for late reply @AKHCE! You should check out https://github.com/MilesCranmer/PySR/discussions/401 for how to do this, which basically finds Note that there is no built-in second-order autodiff, only first-order. (My hope is that Enzyme.jl will eventually give us this for free). A temporary workaround would be to use finite difference approximation for second-order. |
Beta Was this translation helpful? Give feedback.
-
Hi,
I am wondering if anyone has tried to find an analytical approximate solution for a PDE using PySR. I mean, it seems feasible to apply a similar methodology to PINNs for PySR; specifically, in each iteration in PySR, the error of each equation proposed by PySR is computed through the sum of errors associated with the PDE, BC, and IC. I think this approach can be called Physics-Informed PySR. I am curious, have you known/come across any study about Physics-Informed PySR?
Any idea/viewpoint is much appreciated,
Thanks
Beta Was this translation helpful? Give feedback.
All reactions