Train

 

Accueil  Précédent  Suivant

 

 

 

C/C++ declaration

extern long __stdcall Train(long Handle, long* Number, long TrainStyle, TCallBackFct* ACallBack);

 

Pascal declaration

function Train(Handle: LongInt; var Number: Integer; TrainStyle: Integer; ACallBack: TCallBackFct): LongInt; stdcall;

 

Training algorithm has been fixed with SetIValue with style ipBootStrapType.

Launch a training

 

IN                 Handle : Model handle.

IN/OUT         Number: long

 

IN                number or requested epochs

OUT         number of actual epochs

 

IN                TrainStyle: Combination of constants :

                         (0) TrS_Std                Standard training

                         (1) TrS_InitParam          Training with parameters initialization at the first cycle.

                         (2) TrS_Quiet              Quiet training (no info send with ACallBack)

                         (3) TrS_leverage          leverage computation after training.

                         (4) TrS_BootStrap        Bootstrap training from an already trained model.

                     The number of bootstrap is fixed with funxtion SetIValue with style "ipBootStrap".

                     The bootstrap type is fixed with SetIValue with style "ipBootStrapType". Available types are :

                       btStd (1)   : standard bootstrap

                       btResidual (2): residual bootstrap.

IN                ACallBack: Callback function

               

Return         0 if OK, else error code:

-1  invalid handle

 

Remark : Before using this function, the data must be loaded. parameters may be initialized. In orfder to follow the training process, use the callback function and follow the returned values.

 

CreateNewModel

LoadModelFromFile

Primitives


NETRAL Neuro Developer Kit version 7.0