// This file contains the declarations of the classes used in the program

// This class contains a single input data vector as well as a pointer to the
// next data vector in line.
class CDataVector
{
public:
	char									m_output;
	char								 *m_pInputs;

	CDataVector					 *m_pNext;

	CDataVector(void) : m_pInputs(NULL), m_pNext(NULL) {}
	CDataVector(unsigned short numInputs);
	~CDataVector(void);

	CDataVector* Add(unsigned short numInputs);
	CDataVector* Add(char *pBitmapFile, char output);
	CDataVector* getNthVector(unsigned short n);
};

// The node class. It contains the parameters of the neuron as well as
// the input vectors. The vectors are stored as a linked list in the CDataVector
// class. The m_dataVectors never stores any actual vectors, but acts as the
// start of the linked list. It also should be noted that the value stored in
// m_MSE is computed as the network trains (on-line) so m_MSE will be a bit off
// from its real value.
class CNode
{
	unsigned short				m_numInputs;
	unsigned short				m_numEqConnections;
	struct SEqConnection
	{
		double								m_weight;
	}											*m_pEqConnections;

	unsigned long					m_numTuples;
	unsigned short			 *m_pTuplePointers;

	double								m_fracTrainingVectors;
	CDataVector					 *m_pLastTrainingVector;
	unsigned short				m_lastTrainingVector;

	CDataVector						m_dataVectors;
	unsigned short				m_numVectors;

	bool									m_bConverged;
	double								m_MSE;

	unsigned short				m_width;

public:
	static double					learningRate;

	CNode(unsigned short width,
				double fracTrainingVectors,
				char *pWeightsFile);
	~CNode(void);

	bool ReadDataVectors(char *pFileName);
	double ProcessVector(CDataVector *pData);
	bool ProcessEpoch(double MSEgoal);
	inline double run(CDataVector *pData=NULL);
	void print(void);

	bool converged(void)
	{
		return m_bConverged;
	}
	unsigned short getNumVectors(void)
	{
		return m_numVectors;
	}
	double getMSE(void)
	{
		return m_MSE;
	}
};

double CNode::learningRate;