Inheritance diagram for StPmdNeuNet:

Public Member Functions
	StPmdNeuNet ()
	Constructor with no parameter . Purpose ??

	StPmdNeuNet (const Text_t name, Int_t nInput=5, const Text_t hidden="6:7:8", Int_t nOutput=4)
	Constructor.

void	setDiscMaker (StPmdDiscriminatorMaker *)

virtual void	SetKernel (Int_t nInput, const Text_t *hidden, Int_t nOutput)

virtual void	SetLearnParam (Double_t learnParam=0.2, Double_t fse=0., Double_t mu=0.)

virtual void	SetInitParam (Float_t lowerInitWeight=-1., Float_t upperInitWeight=1.)
	Sets the initialisation parameters : max and min weights.

virtual void	Init ()

virtual void	PrintS ()
	prints structure of network on screen

virtual void	Mix ()

virtual Double_t	TrainOneCycle ()

virtual void	ResetCycles ()

virtual void	Export (const Text_t *fileName="exportNN.dat")

virtual void	Import (const Text_t *fileName="exportNN.dat")

virtual void	SetUseBiases (Bool_t trueForUse=1)

virtual void	SetRandomSeed (UInt_t seed=0)

virtual UInt_t	GetRandomSeed ()

virtual Bool_t	IsTrained ()

virtual Int_t	GetNTrainCycles ()

virtual Int_t	GetNTrainEvents ()

virtual void	SetNTrainEvents (Int_t nevt)

virtual Int_t	GetNValidEvents ()

virtual void	SetArraySize (Int_t s=0)

virtual void	FillArray (Int_t, Int_t, Float_t)

virtual void	Fill (Int_t iev=0)

virtual Float_t *	GetInputAdr ()

virtual void	SetInput (Float_t v, Int_t i)

virtual Int_t	GetNInput ()

virtual Int_t	GetNOutput ()

virtual Float_t	GetOutput (Int_t unit=0)

virtual Float_t *	GetOutputAdr ()

virtual Float_t *	GetTeachAdr ()

virtual void	SetTeach (Float_t v, Int_t i)

virtual void	fillArrayOut (Float_t v, Int_t i, Int_t l)

virtual Double_t	GoThrough ()

virtual Float_t	GetSumO ()

void	PrintTrain ()

virtual Double_t	Valid ()

virtual void	TrainNCycles (Int_t nCycles=10)

virtual Int_t	GetNWeights ()

virtual Double_t	ApplyWeights (Float_t , Float_t )

Protected Member Functions
virtual Double_t	Sigmoide (Double_t x)

virtual Double_t	SigPrim (Double_t x)

Protected Attributes
StPmdDiscriminatorMaker *	m_DiscMaker

Detailed Description

Definition at line 54 of file StPmdNeuNet.h.

Member Function Documentation

Double_t StPmdNeuNet::ApplyWeights	(	Float_t *	Teach,
		Float_t *	Value
	)

virtual

one loop on internal events = one cycle. takes each event from internal array in an order fixed by an array ( fEventsList ). It is necessary to call the method Mix() before each call to this function in order to change the presentation order. The learning is done by this function. The private variable fNTrainCycles is incremented.

Definition at line 771 of file StPmdNeuNet.cxx.

void StPmdNeuNet::Export ( const Text_t * fileName = "exportNN.dat" )

virtual

Put the structure in a file WARNING : the weights and biases are stored with 4 digits in decimal part. Learning parameters are not stored

Definition at line 564 of file StPmdNeuNet.cxx.

void StPmdNeuNet::Import ( const Text_t * fileName = "exportNN.dat" )

virtual

Get the structure from a file WARNING : the weights and biases are stored with 4 digits in decimal part. Learning parameteres are not stored.

Definition at line 606 of file StPmdNeuNet.cxx.

void StPmdNeuNet::Init ( )

virtual

initialisation of biases and weights. the init parameters can be changed by : SetInitParam(Float_t lowerInitWeight, Float_t upperInitWeight) The default is -1 and 1

Definition at line 271 of file StPmdNeuNet.cxx.

void StPmdNeuNet::Mix ( )

virtual

mix the events before learning. VERY IMPORTANT. is has to be used before TrainOneCycle() , IT IS NOT used by TrainOneCycle() , you have to do the call yourself

Definition at line 656 of file StPmdNeuNet.cxx.

Referenced by TrainNCycles().

void StPmdNeuNet::SetLearnParam	(	Double_t	learnParam = `0.2`,
		Double_t	fse = `0.`,
		Double_t	mu = `0.`
	)

virtual

Sets the learning parameters : the main learning parameter is around 0.2 (in ]0,1]) fse is for flat spot elimination, with values in [0,0.25], often 0.1 mu is for backprop momentum, values in [0,1]

Definition at line 222 of file StPmdNeuNet.cxx.

void StPmdNeuNet::TrainNCycles ( Int_t nCycles = 10 )

virtual

method to train on N cycles, with mixing and plot of errors on the controller conte.

Definition at line 540 of file StPmdNeuNet.cxx.

References Mix(), TrainOneCycle(), and Valid().

Double_t StPmdNeuNet::TrainOneCycle ( )

virtual

one loop on internal events = one cycle. takes each event from internal array in an order fixed by an array ( fEventsList ). It is necessary to call the method Mix() before each call to this function in order to change the presentation order. The learning is done by this function. The private variable fNTrainCycles is incremented.

Definition at line 477 of file StPmdNeuNet.cxx.

Referenced by TrainNCycles().

Double_t StPmdNeuNet::Valid ( )

virtual

one loop on valid events. takes each event from validation tree. the events are passed trough the kernel, and a mean output error is computed.

Definition at line 518 of file StPmdNeuNet.cxx.

Referenced by TrainNCycles().

The documentation for this class was generated from the following files:

StPmdDiscriminatorMaker/StPmdNeuNet.h
StPmdDiscriminatorMaker/StPmdNeuNet.cxx

Public Member Functions

Protected Member Functions

Protected Attributes

Detailed Description

Member Function Documentation