test1-natural.cpp

neural network 一个演示原理的代码

#include "stdafx.h"
#include "aine.h"
#include "ne-neuron.h"
#include "ne-data.h"

using namespace aine;

extern ofstream tron;

namespace Natural
{
	int TestAdjust( const string )
	{
		Neuron N[10];

		//	Create a very simple neural network net... we have only connection to dendrites.
		//	All threshold values are 1
		//	All dendrite learning ratios are 1
		for( int i = 0 ; i < 4 ; i++ )
			for( int j = 0 ; j < 3 ; j++ )
			{
				N[i].ConnectWith( &N[4 + j] );
				for( int k = 0 ; k < 3 ; k++ )
					N[4 + j].ConnectWith( &N[7 + k] );
			}

		IOListData data1;
		//										INPUT		   |     OUTPUT
		//								  I1    I2    I3    I4 |  O1    O2    O3
		data1.push_back( IOData( 4 , 3 , 1.0 , 0.0 , 0.0 , 0.0 , 0.0 , 1.0 , 0.0 ) );
		data1.push_back( IOData( 4 , 3 , 0.0 , 1.0 , 0.0 , 0.0 , 0.0 , 0.0 , 1.0 ) );
		data1.push_back( IOData( 4 , 3 , 1.0 , 0.0 , 1.0 , 0.0 , 0.0 , 1.0 , 0.0 ) );
		data1.push_back( IOData( 4 , 3 , 0.0 , 0.0 , 0.0 , 1.0 , 1.0 , 0.0 , 0.0 ) );

		bool POK = true;
		const double Epsilon = 0.2;
		int	cnt = 0;

		for( vector<IOData>::iterator it = data1.begin() ; it != data1.end() ; )
		{

			for( int i = 0 ; i < (int)it->input.size() ; i++ )
				N[i].SetInputValue( it->input[i] );

			for( int i = 0 ; i < (int)it->input.size() ; i++ )
				N[i].Depolarization();

			bool IOOK = false;
			
			while( !IOOK )
			{
				IOOK = true;
				for( int i = 0 ; i < (int)it->output.size() ; i++ )
				{
					const double v1 = it->output[i];
					const double v2 = N[4 + i].GetResult();
					if( fabs( v1 - v2 ) > Epsilon )
					{
						N[4 + i].Adjust( it->output[i] );
						IOOK = false;
					}
				}
				cout << ++cnt << ") " << POK << endl;
			}

			for( int i = 0 ; i < 10 ; i++ )
				N[i].Refractory();

			if( !IOOK )
				it = data1.begin();
			else
				++it;
		}
		tron << endl << "after a small adjustment on " << cnt << " turns";
		for( int i = 0 ; i < 7 ; i++ )
			N[i].Trace( &tron );

		tron << endl << "TEST all IO data just for sure..." << endl;
		for( vector<IOData>::iterator it = data1.begin() ; it != data1.end() ; ++it )
		{
			tron << "INPUT: ";
			SaveEach( it->input.begin() , it->input.end() , ostream_iterator<double>( tron , " " ) );

			for( int i = 0 ; i < (int)it->input.size() ; i++ )
				N[i].SetInputValue( it->input[i] );

			for( int i = 0 ; i < (int)it->input.size() ; i++ )
				N[i].Depolarization();

			tron << endl << "    OUTPUT: " << endl;
			for( int i = 0 ; i < (int)it->output.size() ; i++ )
				tron << "\t( " << N[4 + i].GetResult() << " ) " << string( N[4 +i].ActionPotential() ? " T " : " F " ) << " [" << it->output[i] << "] , ";
			tron << endl;

			for( int i = 0 ; i < 10 ; i++ )
				N[i].Refractory();
		}

		return 0;
	}

	int TestInverse( const string )
	{
		Neuron in;
		Neuron inv;

		in.ConnectTo( &inv );
		
		cout << "Tracing NOT:" << endl;
		tron << "Tracing NOT:" << endl;
		for( int i = 0 ; i < 40 ; i++ )
		{
			inv.SetInputValue( 1.0 );
			const double v = (double)((int)Random( 0 , 2 , true ) == 0 ? 0.0 : 1.0);
			in.SetInputValue( v );
			in.Depolarization();
			cout << i << ") " << " NOT " << v << " ---> " << (int)(inv.ActionPotential() ? 1 : 0) << endl;
			tron << i << ") " << " NOT " << v << " ---> " << (int)(inv.ActionPotential() ? 1 : 0);
			if( inv.ActionPotential() )
				if( v == 1.0 )
					tron << " **** ERROR *** ";
			tron << endl;
		}

		return 0;
	}

	int EORTest( const string )
	{
		cout << "create a sample meural Process of EOR Natural method" << endl;
		tron << "create a sample meural Process of EOR Natural method" << endl;

		double in[4][2] =
		{	{	0.0 , 0.0	} ,
			{	1.0 , 0.0	} ,
			{	0.0 , 1.0	} ,
			{	1.0 , 1.0	} 
		};

		Neuron n1 , n2 , inv , kn;

		n1.ConnectWith( &n2 );
		SynapticKnob *sk = n2.ConnectTo( &inv );

		kn.SetInputValue( 1.0 );
		kn.ConnectTo( sk );
		kn.ConnectWith( &kn );

		cout << "Tracing EOR:" << endl;
		tron << "Tracing EOR:" << endl;
		for( int i = 0 ; i < 4 ; i++ )
		{
			const int at = i >= 4 ? (int)Random( 0 , 4 , true ) : i;
			n1.SetInputValue( in[at][0] );
			n2.SetInputValue( in[at][1] );

			cout << i << ") " << in[at][0]  << " EOR " << in[at][1] << " : " ;
			tron << i << ") " << in[at][0]  << " EOR " << in[at][1] << " : " ;

			n1.Depolarization();
			inv.SetInputValue( 1.0 );
			tron << " --> " << " (n2 = " << n2.GetResult() <<" ) --> ";
			n2.Depolarization();
			cout << (int)(inv.ActionPotential() ? 1 : 0) << endl;
			tron << (int)(inv.ActionPotential() ? 1 : 0);
			if( inv.ActionPotential() )
				if( in[at][0] == in[at][1] ) 
					tron << " **** ERROR *** ";
			tron << endl;
		}
		return 0;
	}

	int BoolTest( const string )
	{
		cout << "create a sample meural Process of boolean operators with less neurons and flexible threshold" << endl;
		tron << "\n\ncreate a sample meural Process of boolean operators with less neurons and flexible threshold" << endl;

		Neuron	n1 , n2 , thv;
		double in[4][2] =
		{	{	0.0 , 0.0	} ,
			{	1.0 , 0.0	} ,
			{	0.0 , 1.0	} ,
			{	1.0 , 1.0	} 
		};
		n1.ConnectWith( &n2 );
//		n2.ConnectWith( &n2 );
		thv.ConnectTo( &n2 );
		thv.ConnectWith( &thv );
		cout << "Tracing OR:" << endl;
		tron << "Tracing OR:" << endl;
		for( int i = 0 ; i < 40 ; i++ )
		{
			thv.SetInputValue( 0 );
			thv.Depolarization();

			const int at = i >= 4 ? (int)Random( 0 , 4 , true ) : i;
			n1.SetInputValue( in[at][0] );
			n2.SetInputValue( in[at][1] );

			n1.Depolarization();

			cout << i << ") " << in[at][0]  << " OR " << in[at][1] << " ---> " << (int)(n2.ActionPotential() ? 1 : 0) << endl;
			tron << i << ") " << in[at][0]  << " OR " << in[at][1] << " ---> " << (int)(n2.ActionPotential() ? 1 : 0);
			if( n2.ActionPotential() )
				if( !(in[at][0] || in[at][1]) )
					tron << " **** ERROR *** ";
			tron << endl;
		}
		cout << endl;
		tron << endl;


		cout << "Tracing AND:" << endl;
		tron << "Tracing AND:" << endl;
		for( int i = 0 ; i < 40 ; i++ )
		{
			thv.SetInputValue( 1.0 );
			thv.Depolarization();

			const int at = i >= 4 ? (int)Random( 0 , 4 , true ) : i;
			n1.SetInputValue( in[at][0] );
			n2.SetInputValue( in[at][1] );

			n1.Depolarization();

			cout << i << ") " << in[at][0]  << " AND " << in[at][1] << " ---> " << (int)(n2.ActionPotential() ? 1 : 0) << endl;
			tron << i << ") " << in[at][0]  << " AND " << in[at][1] << " ---> " << (int)(n2.ActionPotential() ? 1 : 0);
			if( n2.ActionPotential() )
				if( !(in[at][0] && in[at][1]) )
					tron << " **** ERROR *** ";
			tron << endl;
		}
		cout << endl;
		tron << endl;

		Neuron inv;
		n2.ConnectTo( &inv );

		cout << "Tracing NOR:" << endl;
		tron << "Tracing NOR:" << endl;
		for( int i = 0 ; i < 40 ; i++ )
		{
			thv.SetInputValue( 0 );
			thv.Depolarization();

			const int at = i >= 4 ? (int)Random( 0 , 4 , true ) : i;
			n1.SetInputValue( in[at][0] );
			n2.SetInputValue( in[at][1] );

			n1.Depolarization();

			inv.SetInputValue( 1.0 );
			n2.Depolarization();

			cout << i << ") " << in[at][0]  << " NOR " << in[at][1] << " ---> " << (int)(inv.ActionPotential() ? 1 : 0) << endl;
			tron << i << ") " << in[at][0]  << " NOR " << in[at][1] << " ---> " << (int)(inv.ActionPotential() ? 1 : 0);
			if( inv.ActionPotential() )
				if( in[at][0] || in[at][1] )
					tron << " **** ERROR *** ";
			tron << endl;
			
		}
		cout << endl;
		tron << endl;

		cout << "Tracing NAND:" << endl;
		tron << "Tracing NAND:" << endl;
		for( int i = 0 ; i < 40 ; i++ )
		{

			thv.SetInputValue( 1.0 );
			thv.Depolarization();
			const int at = i >= 4 ? (int)Random( 0 , 4 , true ) : i;
			n1.SetInputValue( in[at][0] );
			n2.SetInputValue( in[at][1] );

			n1.Depolarization();

			cout << i << ") " << in[at][0]  << " AND " << in[at][1] << " ---> " << (int)(n2.ActionPotential() ? 1 : 0) << endl;
			tron << i << ") " << in[at][0]  << " AND " << in[at][1] << " ---> " << (int)(n2.ActionPotential() ? 1 : 0);
			if( n2.ActionPotential() )
				if( !(in[at][0] && in[at][1]) )
					tron << " **** ERROR *** ";
			tron << "\t";

			inv.SetInputValue( 1.0 );
			n2.Depolarization();

			cout << i << ") " << in[at][0]  << " NAND " << in[at][1] << " ---> " << (int)(inv.ActionPotential() ? 1 : 0) << endl;
			tron << i << ") " << in[at][0]  << " NAND " << in[at][1] << " ---> " << (int)(inv.ActionPotential() ? 1 : 0);
			if( inv.ActionPotential() )
				if( (in[at][0] && in[at][1]) )
					tron << " **** ERROR *** ";
			tron << endl;
			
		}
		cout << endl;
		tron << endl;

		return 0;
	}
}