ENH: new hybrid hydrological models using neural network and ODE-based system

DassHydro / smash

An open source, Python library interfacing the Fortran Spatially distributed Modelling and ASsimilation for Hydrology platform.

https://smash.recover.inrae.fr/

GNU General Public License v3.0

12 stars 6 forks source link

ENH: new hybrid hydrological models using neural network and ODE-based system #294

Closed nghi-truyen closed 1 month ago

nghi-truyen commented 3 months ago

New hydrological models added to smash:

gr4_mlp : integrated-neural network GR4 model for correcting internal fluxes in algebraic scheme.
gr4_ode : ODE-based GR4 model
gr4_ode_mlp : integrated-neural network GR4 model for correcting internal fluxes in ODE scheme.

TODO: the examples in the documentation of baysian_optimize and baysian_optimize_control_info will be corrected in a later PR since they will be modified (when adding adaptive optimizers to all mappings).

inoelloc commented 3 months ago

107 commits well done !

nghi-truyen commented 3 months ago

@inoelloc Voilà, le nouveau commit avec toutes les choses qu'on a discutées. Le seul endroit qui change le baseline est le truc sur l'IO de model_ddt (à cause de l'ajout de weight_3 + bias_3 et la suppression de hidden_neuron que je pense être un peu redondant dans save_model_ddt).