wuuu

Nerve.sln

@@ -7,6 +7,8 @@ Project("{8BC9CEB8-8B4A-11D0-8D11-00A0C91BC942}") = "Nerve", "Nerve\Nerve.vcxpro
 EndProject
 Project("{8BC9CEB8-8B4A-11D0-8D11-00A0C91BC942}") = "Reform_TrainingData", "Reform_TrainingData\Reform_TrainingData.vcxproj", "{EF1B4ECE-FA22-49BD-B4C9-660088459C51}"
 EndProject
+Project("{8BC9CEB8-8B4A-11D0-8D11-00A0C91BC942}") = "Train", "Train\Train.vcxproj", "{CC43A4BC-E8D8-425B-954B-3B98D4AE6155}"
+EndProject
 Global
 	GlobalSection(SolutionConfigurationPlatforms) = preSolution
 		Debug|x64 = Debug|x64
@@ -31,6 +33,14 @@ Global
 		{EF1B4ECE-FA22-49BD-B4C9-660088459C51}.Release|x64.Build.0 = Release|x64
 		{EF1B4ECE-FA22-49BD-B4C9-660088459C51}.Release|x86.ActiveCfg = Release|Win32
 		{EF1B4ECE-FA22-49BD-B4C9-660088459C51}.Release|x86.Build.0 = Release|Win32
+		{CC43A4BC-E8D8-425B-954B-3B98D4AE6155}.Debug|x64.ActiveCfg = Debug|x64
+		{CC43A4BC-E8D8-425B-954B-3B98D4AE6155}.Debug|x64.Build.0 = Debug|x64
+		{CC43A4BC-E8D8-425B-954B-3B98D4AE6155}.Debug|x86.ActiveCfg = Debug|Win32
+		{CC43A4BC-E8D8-425B-954B-3B98D4AE6155}.Debug|x86.Build.0 = Debug|Win32
+		{CC43A4BC-E8D8-425B-954B-3B98D4AE6155}.Release|x64.ActiveCfg = Release|x64
+		{CC43A4BC-E8D8-425B-954B-3B98D4AE6155}.Release|x64.Build.0 = Release|x64
+		{CC43A4BC-E8D8-425B-954B-3B98D4AE6155}.Release|x86.ActiveCfg = Release|Win32
+		{CC43A4BC-E8D8-425B-954B-3B98D4AE6155}.Release|x86.Build.0 = Release|Win32
 	EndGlobalSection
 	GlobalSection(SolutionProperties) = preSolution
 		HideSolutionNode = FALSE

Nerve/Nerve.cpp

@@ -274,56 +274,56 @@ namespace Nerve
 }
 using namespace Nerve;
 
-int main()
-{
-	//Nerve_net net(4, 3, vector<int>{3,4});
-	Nerve_net net("a.nerve");//从文件读入
-
-	int o = 30000;
-	default_random_engine e(GetTickCount());/*初始化随机引擎*/
-	uniform_int_distribution<int> u(0,2);
-	while (o--)
-	{
-		if (u(e) == 0)
-		{
-			net.Input(vector<double>{1, 1, 0, 0});
-			net.Set_Desired_output(vector<double>{1, 0, 0});
-			net.Figue();
-			net.Train();
-		}
-		else if (u(e) == 1)
-		{
-			net.Input(vector<double>{0, 1, 1, 0});
-			net.Set_Desired_output(vector<double>{0, 1, 0});
-			net.Figue();
-			net.Train();
-		}
-		else
-		{
-			net.Input(vector<double>{0, 0, 1, 1});
-			net.Set_Desired_output(vector<double>{0, 0, 1});
-			net.Figue();
-			net.Train();
-		}
-
-		net.Input(vector<double>{1, 1, 1, 1});
-		net.Set_Desired_output(vector<double>{1, 1, 1});
-		net.Figue();
-		net.Train();
-
-	}
-	//教会它识别01
-	net.Input(vector<double>{1,0.5,0,0.0});
-	net.Figue();
-	vector<double> oo = net.Output();
-	for (int i = 0; i < oo.size(); i++)
-		std::cout << oo[i] << " ";
-	std::cout << endl;
-	net.Input(vector<double>{0,0.60,1.0,0.00});
-	net.Figue();
-	vector<double> ooo = net.Output();
-	for (int i = 0; i < ooo.size(); i++)
-		std::cout << ooo[i] << " ";
-
-	//net.SaveTofile("a.nerve");//保存到文件
-}
+//int main()
+//{
+//	//Nerve_net net(4, 3, vector<int>{3,4});
+//	Nerve_net net("a.nerve");//从文件读入
+//	
+//	int o = 30000;
+//	default_random_engine e(GetTickCount());/*初始化随机引擎*/
+//	uniform_int_distribution<int> u(0,2);
+//	while (o--)
+//	{
+//		if (u(e) == 0)
+//		{
+//			net.Input(vector<double>{1, 1, 0, 0});
+//			net.Set_Desired_output(vector<double>{1, 0, 0});
+//			net.Figue();
+//			net.Train();
+//		}
+//		else if (u(e) == 1)
+//		{
+//			net.Input(vector<double>{0, 1, 1, 0});
+//			net.Set_Desired_output(vector<double>{0, 1, 0});
+//			net.Figue();
+//			net.Train();
+//		}
+//		else
+//		{
+//			net.Input(vector<double>{0, 0, 1, 1});
+//			net.Set_Desired_output(vector<double>{0, 0, 1});
+//			net.Figue();
+//			net.Train();
+//		}
+//
+//		net.Input(vector<double>{1, 1, 1, 1});
+//		net.Set_Desired_output(vector<double>{1, 1, 1});
+//		net.Figue();
+//		net.Train();
+//
+//	}
+//	//教会它识别01
+//	net.Input(vector<double>{1,0.5,0,0.0});
+//	net.Figue();
+//	vector<double> oo = net.Output();
+//	for (int i = 0; i < oo.size(); i++)
+//		std::cout << oo[i] << " ";
+//	std::cout << endl;
+//	net.Input(vector<double>{0,0.60,1.0,0.00});
+//	net.Figue();
+//	vector<double> ooo = net.Output();
+//	for (int i = 0; i < ooo.size(); i++)
+//		std::cout << ooo[i] << " ";
+//
+//	//net.SaveTofile("a.nerve");//保存到文件
+//}

Reform_TrainingData/reform.cpp

@@ -63,16 +63,20 @@ class Builder
 
 		bool end = false;
 		int count = 0;
+		int lis[10] = { 0,1,2,3,4,5,6,7,8,9 };
 		while (!end)
 		{
+			for (int i = 0; i<10; i++)//随机1-9的输出顺序 
+				swap(lis[i], lis[rand() % 10]);
+			end = true;
 			for (int i = 0; i < 10; i++)
 			{
-                end=true;
-				if(count<data[i].size())
-					out.write((char*)&data[i][count], sizeof(Data)),end=false;
+				if (count<data[lis[i]].size())
+					out.write((char*)&data[lis[i]][count], sizeof(Data)), end = false;
 			}
 			count++;
 		}
+		out.close();
 	}
 private:
 	ifstream	in;
@@ -93,8 +97,8 @@ class Builder
 };
 int main()
 {
-    Builder r("train-images.idx3-ubyte","train-labels.idx1-ubyte");
+    Builder r("t10k-images.idx3-ubyte","t10k-labels.idx1-ubyte");
     r.wan();
-    r.Export("export.txt");
+    r.Export("export.exam");
     return 0;
 }

Reform_TrainingData/x64/Release/Reform_TrainingData.log

@@ -1,4 +1,9 @@
   reform.cpp
 reform.cpp(54): warning C4267: “参数”: 从“size_t”转换到“int”，可能丢失数据
-MSVCRT.lib(exe_main.obj) : error LNK2001: 无法解析的外部符号 main
-E:\我的 东东\EasyX学习\深度学习\Nerve_Net\x64\Release\Reform_TrainingData.exe : fatal error LNK1120: 1 个无法解析的外部命令
+  正在生成代码
+  4 of 236 functions ( 1.7%) were compiled, the rest were copied from previous compilation.
+    0 functions were new in current compilation
+    4 functions had inline decision re-evaluated but remain unchanged
+  已完成代码的生成
+  Reform_TrainingData.vcxproj -> E:\我的 东东\EasyX学习\深度学习\Nerve_Net\x64\Release\Reform_TrainingData.exe
+  Reform_TrainingData.vcxproj -> E:\我的 东东\EasyX学习\深度学习\Nerve_Net\x64\Release\Reform_TrainingData.pdb (Full PDB)

Train/Train.cpp

@@ -0,0 +1,91 @@
+#include<iostream>
+#include<fstream>
+#include<vector>
+#include"..//Nerve//Nerve.cpp"
+#include<windows.h>
+using namespace std;
+/*
+单体数据结构
+*/
+#define N 28
+struct Train
+{
+	int train_times;	//训练次数
+
+	int exam_times;		//测试次数 
+
+	int exam_A;			//正确样本数
+	int exam_W;			//错误样本数 
+
+	Nerve_net &nerve;	//训练神经网络 
+	char Nerve_ID[16];	//神经网络ID 
+	struct Data
+	{
+		char point[N * N];
+		char ans;
+	};
+
+public:
+	Train(Nerve_net &i):nerve(i)
+	{
+	}
+	void train(char *file)
+	{
+		ifstream in(file,ios::binary);
+		Data data;
+		vector<double> vec(N*N,0.0);
+		vector<double> outp(10,0.0);
+		while(1)
+		{
+			in.read((char*)&data,sizeof(data));
+			if(in.gcount()!=0)
+			{
+				for(int i=0;i<N*N;i++)
+					vec[i]=data.point[i];
+				outp[data.ans]=1.0;
+				nerve.Input(vec);
+				nerve.Set_Desired_output(outp);
+				nerve.Figue();
+				nerve.Train();
+
+				train_times++;
+			}
+			else break;
+		}
+	}
+	void exam(char *file)
+	{
+		ifstream in(file,ios::binary);
+		Data data;
+		vector<double> vec(N*N,0.0);
+		vector<double> ans(10,0.0);
+		while(1)
+		{
+			in.read((char*)&data,sizeof(data));
+			if(in.gcount()!=0)
+			{
+				for(int i=0;i<N*N;i++)
+					vec[i]=data.point[i];
+				nerve.Input(vec);
+				nerve.Figue();
+				ans=nerve.Output();
+				double maxx=0;
+				int k=0;
+				for(int i=0;i<10;i++)
+				if(maxx<ans[i])
+					maxx=ans[i],k=i;
+				if(k==data.ans)exam_A++;
+				else exam_W++;
+			}
+			else break;
+		}
+	}
+	void output_statistic(ostream &out)
+	{
+	}
+};
+int main()
+{
+
+    return 0;
+}