search

PhIMaL / DeePyMoD

https://phimal.github.io/DeePyMoD/
MIT License
44 stars 11 forks source link

some wrong results for ode #70

Open xgxg1314 opened 2 years ago

xgxg1314 commented 2 years ago

https://phimal.github.io/DeePyMoD/examples/ODE_nonlin/ODE/

In this ode example, The set of ODEs considered here are d[y, z]/dt = [z, -z+ 5 \sin y], but we can get some wrong identified coefficients using DeePyMod. I have no idea what happened. Could you please help me?

[array([[0. ], [0. ], [0.99987924], [0. ]], dtype=float32), array([[ 0. ], [ 0. ], [-0.56510067], [-1.076641 ]], dtype=float32)]

georgemilosh commented 11 months ago

Also for me it is not clear. Even if the input is already normalized

georgemilosh commented 11 months ago

Ok, so after some tinkering around I report that the physical coefficients are stored in model.constraint_coeffs() rather than model.estimator_coeffs().