C++感知机分类Iris数据集

最新推荐文章于 2024-04-29 13:51:30 发布
最新推荐文章于 2024-04-29 13:51:30 发布
阅读量602 收藏 2
点赞数
CC 4.0 BY-SA版权
分类专栏： 神经网络 文章标签： 感知机 机器学习 神经网络
版权声明：本文为博主原创文章，遵循 CC 4.0 BY-SA 版权协议，转载请附上原文出处链接和本声明。
本文链接：https://blog.youkuaiyun.com/yongchou3909/article/details/79157281

目前为止尝试了两种办法，因为是个四维数据要分三类：

想看好结果直接移步第二种方法（Queens 2017 meng的同学如果搜索到这篇文章请不要copy代码谢谢）

有更好办法欢迎评论，觉得我太垃圾直接右上角点X

第一种方法：设置三个权向量w1,w2,w3，第一类（1,0,0），第二类（0,1,0）第三类（0,0,1），1表示大于零0表示小于零，训练每个数据集的时候计算y1y2y3然后看正负。输入自己不会写是贴的别人代码。。很惭愧了，训练大概20000词的时候能得到最佳结果，50%的准确率，可见这个方法不好。训练10次以内w1稳定，准确率100%，20000次左右w3稳定，w2在40000次还不稳定。

代码：

#include<iostream>
#include<fstream>
#include<vector>
#include<math.h>
#include<string>
#include<sstream>
#include<stdlib.h>
#include<time.h>
#include<string.h>
using namespace std;


struct Tuple {
string attr1;
string attr2;
string attr3;
string attr4;
string attr5;
};


struct Tuple1 {
float attr1;
float attr2;
float attr3;
float attr4;
string attr5;
};
double weight1[4] = { 0, 0, 0, 0 };
double weight2[4] = { 0, 0, 0, 0 };
double weight3[4] = { 0, 0, 0, 0 };
double bias = 0;
double bias2 = 0;
double bias3 = 0;
int count = 0;
double y = 0;
double y2 = 0;
double y3 = 0;
int error;
int t;
int rep;
int main() {
char fname[255] = "C:\\Users\\Administrator\\Desktop\\train.txt";
ifstream infile;
infile.open(fname, ios::in);
if (!infile) {
cout << "Can not read the training file" << endl;
while (1)
;
return 0;
}
vector<Tuple> tuples;
vector<Tuple1> tuples1;
Tuple tuple;
Tuple1 tuple1;
int remainder = 0;
//Reading data from train.txt
while (!infile.eof()) {


count++;
remainder = count % 5;


switch (remainder) {
case 0:
getline(infile, tuple.attr5, '\n');
tuples.push_back(tuple);
break;
case 1:
getline(infile, tuple.attr1, ',');
break;
case 2:
getline(infile, tuple.attr2, ',');
break;
case 3:
getline(infile, tuple.attr3, ',');
break;
case 4:
getline(infile, tuple.attr4, ',');
break;
}
}


for (vector<Tuple>::size_type i = 0; i != tuples.size(); i++) {


tuple1.attr1 = stof(tuples[i].attr1);
tuple1.attr2 = stof(tuples[i].attr2);
tuple1.attr3 = stof(tuples[i].attr3);
tuple1.attr4 = stof(tuples[i].attr4);
tuple1.attr5 = tuples[i].attr5;
tuples1.push_back(tuple1);
}


//Output the data in train.txt
/*for (vector<Tuple1>::size_type i = 0; i != tuples1.size(); ++i) {
cout << tuples1[i].attr1 <<"," <<tuples1[i].attr2<<","  << tuples1[i].attr3 << "," << tuples1[i].attr4 << "," << tuples1[i].attr5 << endl;
}
*/
for (t = 1; t < 40000; t++) {
error = 0;
for (vector<Tuple1>::size_type i = 0; i != tuples1.size(); ++i) {
//y = 0;
//y2 = 0;
//y3 = 0;
y = weight1[0] * tuples1[i].attr1 + weight1[1] * tuples1[i].attr2
+ weight1[2] * tuples1[i].attr3
+ weight1[3] * tuples1[i].attr4 + bias;
y2 = weight2[0] * tuples1[i].attr1 + weight2[1] * tuples1[i].attr2
+ weight2[2] * tuples1[i].attr3
+ weight2[3] * tuples1[i].attr4 + bias2;
y3 = weight3[0] * tuples1[i].attr1 + weight3[1] * tuples1[i].attr2
+ weight3[2] * tuples1[i].attr3
+ weight3[3] * tuples1[i].attr4 + bias3;
if (tuples1[i].attr5 == "Iris-setosa") { //判断是不是第一种花
if (y > 0) {
} //判断第一个权值是否合适
else {
weight1[0] = weight1[0] + 0.1 * (tuples1[i].attr1);
weight1[1] = weight1[1] + 0.1 * (tuples1[i].attr2);
weight1[2] = weight1[2] + 0.1 * (tuples1[i].attr3);
weight1[3] = weight1[3] + 0.1 * (tuples1[i].attr4);
bias = bias + 0.1;
}
if (y2 > 0) { //判断第二个权值是否合适
weight2[0] = weight2[0] - 0.2 * (tuples1[i].attr1);
weight2[1] = weight2[1] - 0.2* (tuples1[i].attr2);
weight2[2] = weight2[2] - 0.2* (tuples1[i].attr3);
weight2[3] = weight2[3] - 0.2 * (tuples1[i].attr4);
bias2 = bias2 - 0.2;
} else {
};
if (y3 > 0) { //判断第三个权值是否合适
weight3[0] = weight3[0] - 0.2* (tuples1[i].attr1);
weight3[1] = weight3[1] - 0.2* (tuples1[i].attr2);
weight3[2] = weight3[2] - 0.2* (tuples1[i].attr3);
weight3[3] = weight3[3] - 0.2* (tuples1[i].attr4);
bias3 = bias3 - 0.2;
} else {
};
cout << i << "flower" << endl;
cout << "weight1 is [" << weight1[0] << "," << weight1[1] << ","
<< weight1[2] << "," << weight1[3] << "]" << endl;
cout << "weight2 is [" << weight2[0] << "," << weight2[1] << ","
<< weight2[2] << "," << weight2[3] << "]" << endl;
cout << "weight3 is [" << weight3[0] << "," << weight3[1] << ","
<< weight3[2] << "," << weight3[3] << "]" << endl;
cout << y << "," << y2 << "," << y3 << endl;
cout << "bias:" << bias << "," << bias2 << "," << bias3 << endl;
} else if (tuples1[i].attr5 == "Iris-versicolor") { //判断是不是第二种花
if (y > 0) { //判断第一个权值是否合适
weight1[0] = weight1[0] - 0.1 * (tuples1[i].attr1);
weight1[1] = weight1[1] - 0.1 * (tuples1[i].attr2);
weight1[2] = weight1[2] - 0.1 * (tuples1[i].attr3);
weight1[3] = weight1[3] - 0.1 * (tuples1[i].attr4);
bias = bias - 0.1;
} else {
}
if (y2 > 0) {
} //判断第二个权值是否合适
else {
weight2[0] = weight2[0] + 0.2 * (tuples1[i].attr1);
weight2[1] = weight2[1] + 0.2* (tuples1[i].attr2);
weight2[2] = weight2[2] + 0.2 * (tuples1[i].attr3);
weight2[3] = weight2[3] + 0.2* (tuples1[i].attr4);
bias2 = bias2 + 0.2;
}
if (y3 > 0) { //判断第三个权值是否合适
weight3[0] = weight3[0] - 0.2 * (tuples1[i].attr1);
weight3[1] = weight3[1] - 0.2 * (tuples1[i].attr2);
weight3[2] = weight3[2] - 0.2 * (tuples1[i].attr3);
weight3[3] = weight3[3] - 0.2 * (tuples1[i].attr4);
bias3 = bias3 - 0.2;
} else {
}
cout << i << "flower" << endl;
cout << "weight1 is [" << weight1[0] << "," << weight1[1] << ","
<< weight1[2] << "," << weight1[3] << "]" << endl;
cout << "weight2 is [" << weight2[0] << "," << weight2[1] << ","
<< weight2[2] << "," << weight2[3] << "]" << endl;
cout << "weight3 is [" << weight3[0] << "," << weight3[1] << ","
<< weight3[2] << "," << weight3[3] << "]" << endl;
cout << y << "," << y2 << "," << y3 << endl;
cout << "bias:" << bias << "," << bias2 << "," << bias3 << endl;
} else { //否则就是第三种花
if (y > 0) { //判断第一个权值是否合适
weight1[0] = weight1[0] - 0.1 * (tuples1[i].attr1);
weight1[1] = weight1[1] - 0.1 * (tuples1[i].attr2);
weight1[2] = weight1[2] - 0.1 * (tuples1[i].attr3);
weight1[3] = weight1[3] - 0.1 * (tuples1[i].attr4);
bias = bias - 0.1;
} else {
}
if (y2 > 0) { //判断第二个权值是否合适
weight2[0] = weight2[0] - 0.2 * (tuples1[i].attr1);
weight2[1] = weight2[1] - 0.2 * (tuples1[i].attr2);
weight2[2] = weight2[2] - 0.2 * (tuples1[i].attr3);
weight2[3] = weight2[3] -0.2 * (tuples1[i].attr4);
bias2 = bias2 - 0.2;
} else {
}
if (y3 > 0) {
} //判断第三个权值是否合适
else {
weight3[0] = weight3[0] + 0.2* (tuples1[i].attr1);
weight3[1] = weight3[1] + 0.2 * (tuples1[i].attr2);
weight3[2] = weight3[2] + 0.2 * (tuples1[i].attr3);
weight3[3] = weight3[3] + 0.2 * (tuples1[i].attr4);
bias3 = bias3 + 0.2;


}
cout << i << "flower" << endl;
cout << "weight1 is [" << weight1[0] << "," << weight1[1] << ","
<< weight1[2] << "," << weight1[3] << "]" << endl;
cout << "weight2 is [" << weight2[0] << "," << weight2[1] << ","
<< weight2[2] << "," << weight2[3] << "]" << endl;
cout << "weight3 is [" << weight3[0] << "," << weight3[1] << ","
<< weight3[2] << "," << weight3[3] << "]" << endl;
cout << y << "," << y2 << "," << y3 << endl;
cout << "bias:" << bias << "," << bias2 << "," << bias3 << endl;
}


}
//cout<<"weight1 is ["<<weight1[0]<<","<<weight1[1]<<","<<weight1[2]<<","<<weight1[3]<<"]"<<endl;
//cout<<"weight2 is ["<<weight2[0]<<","<<weight2[1]<<","<<weight2[2]<<","<<weight2[3]<<"]"<<endl;
//cout<<"weight3 is ["<<weight3[0]<<","<<weight3[1]<<","<<weight3[2]<<","<<weight3[3]<<"]"<<endl;
//cout<<"this is the "<<t<<" time training"<<endl;
for (vector<Tuple1>::size_type i = 0; i != tuples1.size(); ++i) { //计算MSE决定何时停止训练
//error=0;
y = weight1[0] * tuples1[i].attr1 + weight1[1] * tuples1[i].attr2
+ weight1[2] * tuples1[i].attr3
+ weight1[3] * tuples1[i].attr4 + bias;
y2 = weight2[0] * tuples1[i].attr1 + weight2[1] * tuples1[i].attr2
+ weight2[2] * tuples1[i].attr3
+ weight2[3] * tuples1[i].attr4 + bias2;
y3 = weight3[0] * tuples1[i].attr1 + weight3[1] * tuples1[i].attr2
+ weight3[2] * tuples1[i].attr3
+ weight3[3] * tuples1[i].attr4 + bias3;
if (tuples1[i].attr5 == "Iris-setosa") {
if (y > 0) {
} else {
error = error + (-y);
}
if (y2 > 0) {
error = error + y2;


}
}
if (tuples1[i].attr5 == "Iris-versicolor") {


if (y2 > 0) {
} else {
error = error + (-y2);
}


} else {
if (y2 > 0) {
error = error + y2;


}
if (y3 > 0) {
} else {
error = error + (-y3);
}
}


}


cout << "error is " << error << endl;
cout << "this is " << t << "time" << endl;


if (error < 1000 && t != 1) {
//break;
}
}


}
最后的误差计算还没弄懂，主要是计算权值的那点比较关键。测试重新开了一堆代码，从这里面得到结果后把测试代码的权值改了就行。
#include<iostream>
#include<fstream>
#include<vector>
#include<math.h>
#include<string>
#include<sstream>
#include<stdlib.h>
#include<time.h>
#include<string.h>
using namespace std;


struct Tuple {
string attr1;
string attr2;
string attr3;
string attr4;
string attr5;
};


struct Tuple1 {
float attr1;
float attr2;
float attr3;
float attr4;
string attr5;
};
double weight1[4] = {0.06, 0.52, -0.81, -0.34 };
double weight2[4] = {14.3977,-11.5198, -0.501327, -30.6595 };
double weight3[4] = {-7.3401,-56.82, 17.6799, 114.3 };
double bias = 0.1;
double bias2 = -21.8;
double bias3 = -75.6;
int count = 0;
double y = 0;
double y2 = 0;
double y3 = 0;
int error = 0;
int t;
int rep;
int main() {
char fname[255] = "C:\\Users\\Administrator\\Desktop\\test.txt";
ifstream infile;
infile.open(fname, ios::in);
if (!infile) {
cout << "Can not read the test file" << endl;
while (1)
;
return 0;
}
vector<Tuple> tuples;
vector<Tuple1> tuples1;
Tuple tuple;
Tuple1 tuple1;
int remainder = 0;
//Reading data from train.txt
while (!infile.eof()) {


count++;
remainder = count % 5;


switch (remainder) {
case 0:
getline(infile, tuple.attr5, '\n');
tuples.push_back(tuple);
break;
case 1:
getline(infile, tuple.attr1, ',');
break;
case 2:
getline(infile, tuple.attr2, ',');
break;
case 3:
getline(infile, tuple.attr3, ',');
break;
case 4:
getline(infile, tuple.attr4, ',');
break;
}
}


for (vector<Tuple>::size_type i = 0; i != tuples.size(); i++) {


tuple1.attr1 = stof(tuples[i].attr1);
tuple1.attr2 = stof(tuples[i].attr2);
tuple1.attr3 = stof(tuples[i].attr3);
tuple1.attr4 = stof(tuples[i].attr4);
tuple1.attr5 = tuples[i].attr5;
tuples1.push_back(tuple1);
}


//Output the data in train.txt
/*for (vector<Tuple1>::size_type i = 0; i != tuples1.size(); ++i) {
cout << tuples1[i].attr1 <<"," <<tuples1[i].attr2<<","  << tuples1[i].attr3 << "," << tuples1[i].attr4 << "," << tuples1[i].attr5 << endl;
}
*/
for (vector<Tuple1>::size_type i = 0; i != tuples1.size(); ++i) {
y = weight1[0] * tuples1[i].attr1 + weight1[1] * tuples1[i].attr2
+ weight1[2] * tuples1[i].attr3
+ weight1[3] * tuples1[i].attr4 + bias;
y2 = weight2[0] * tuples1[i].attr1 + weight2[1] * tuples1[i].attr2
+ weight2[2] * tuples1[i].attr3
+ weight2[3] * tuples1[i].attr4 + bias2;
y3 = weight3[0] * tuples1[i].attr1 + weight3[1] * tuples1[i].attr2
+ weight3[2] * tuples1[i].attr3
+ weight3[3] * tuples1[i].attr4 + bias3;
//cout <<i<<","<<y << "," << y2 << "," << y3 << endl;


if (y > 0&&y2<0&&y3<0 ) {
cout << i<<",Iris-setosa" << endl;
} else if ( y < 0 &&y2>0&&y3<0) {
cout <<i<< ",Iris-versicolor" << endl;
} else if ( y< 0&&y2<0&&y3>0) {
cout << i<<",Iris-virginica" << endl;
} else {
cout << i<<",Failed" << endl;
}


}
}

第二种方法表现比较好，只用了两个权向量，先用一个向量区分第一类和第三类，再在这个基础上缩小样本空间用另一个向量区分第二类和第三类，训练次数为5352，两个权值均稳定，测试集有5个错误，准确率83%，代码是在第一个的基础上改的，有很多注释，用的话可以直接删掉。

代码：

#include<iostream>
#include<fstream>
#include<vector>
#include<math.h>
#include<string>
#include<sstream>
#include<stdlib.h>
#include<time.h>
#include<string.h>
using namespace std;


struct Tuple {
string attr1;
string attr2;
string attr3;
string attr4;
string attr5;
};


struct Tuple1 {
float attr1;
float attr2;
float attr3;
float attr4;
string attr5;
};
double weight1[4] = { 0, 0, 0, 0 };
double weight2[4] = { 0, 0, 0, 0 };
double weight1tep[4] = { 0, 0, 0, 0 };
double weight2tep[4] = { 0, 0, 0, 0 };
double bias = 0;
double bias2 = 0;
double bias3 = 0;
int count = 0;
double y = 0;
double y2 = 0;
double y3 = 0;
int error;
int t = 1;
int rep;
int m;
int main() {
char fname[255] = "C:\\Users\\Administrator\\Desktop\\train.txt";//read data and input into tuple1
ifstream infile;
infile.open(fname, ios::in);
if (!infile) {
cout << "Can not read the training file" << endl;
while (1)
;
return 0;
}
vector<Tuple> tuples;
vector<Tuple1> tuples1;
Tuple tuple;
Tuple1 tuple1;
int remainder = 0;
//Reading data from train.txt
while (!infile.eof()) {


count++;
remainder = count % 5;


switch (remainder) {
case 0:
getline(infile, tuple.attr5, '\n');
tuples.push_back(tuple);
break;
case 1:
getline(infile, tuple.attr1, ',');
break;
case 2:
getline(infile, tuple.attr2, ',');
break;
case 3:
getline(infile, tuple.attr3, ',');
break;
case 4:
getline(infile, tuple.attr4, ',');
break;
}
}


for (vector<Tuple>::size_type i = 0; i != tuples.size(); i++) {


tuple1.attr1 = stof(tuples[i].attr1);
tuple1.attr2 = stof(tuples[i].attr2);
tuple1.attr3 = stof(tuples[i].attr3);
tuple1.attr4 = stof(tuples[i].attr4);
tuple1.attr5 = tuples[i].attr5;
tuples1.push_back(tuple1);
}//fining reading data and ready to train


//Output the red data
/*for (vector<Tuple1>::size_type i = 0; i != tuples1.size(); ++i) {
cout << tuples1[i].attr1 <<"," <<tuples1[i].attr2<<","  << tuples1[i].attr3 << "," << tuples1[i].attr4 << "," << tuples1[i].attr5 << endl;
}
*/
for (t = 1; t < 20001; t++) {//loop of training


for (m = 0; m < 4; m++) {//this is for recoding the weight value, if the weight never change, break loop


weight1tep[m] = weight1[m];
weight2tep[m] = weight2[m];
}
for (vector<Tuple1>::size_type i = 0; i != tuples1.size(); ++i) {//starting retrive the training data
//y = 0;
//y2 = 0;
//y3 = 0;
y = weight1[0] * tuples1[i].attr1 + weight1[1] * tuples1[i].attr2
+ weight1[2] * tuples1[i].attr3
+ weight1[3] * tuples1[i].attr4 + bias;
y2 = weight2[0] * tuples1[i].attr1 + weight2[1] * tuples1[i].attr2
+ weight2[2] * tuples1[i].attr3
+ weight2[3] * tuples1[i].attr4 + bias2;
/*y3 = weight3[0] * tuples1[i].attr1 + weight3[1] * tuples1[i].attr2
+ weight3[2] * tuples1[i].attr3
+ weight3[3] * tuples1[i].attr4 + bias3;*/
if (tuples1[i].attr5 == "Iris-setosa") { //read the lable of trained data
if (y > 0) {
}//judge and adjust weight1
else {
weight1[0] = weight1[0] + 0.1 * (tuples1[i].attr1);
weight1[1] = weight1[1] + 0.1 * (tuples1[i].attr2);
weight1[2] = weight1[2] + 0.1 * (tuples1[i].attr3);
weight1[3] = weight1[3] + 0.1 * (tuples1[i].attr4);
bias = bias + 0.1;
}
/*if (y2 > 0) { //判断第二个权值是否合适
weight2[0] = weight2[0] - 0.2 * (tuples1[i].attr1);
weight2[1] = weight2[1] - 0.2* (tuples1[i].attr2);
weight2[2] = weight2[2] - 0.2* (tuples1[i].attr3);
weight2[3] = weight2[3] - 0.2 * (tuples1[i].attr4);
bias2 = bias2 - 0.2;
} else {
};
if (y3 > 0) { //判断第三个权值是否合适
weight3[0] = weight3[0] - 0.2* (tuples1[i].attr1);
weight3[1] = weight3[1] - 0.2* (tuples1[i].attr2);
weight3[2] = weight3[2] - 0.2* (tuples1[i].attr3);
weight3[3] = weight3[3] - 0.2* (tuples1[i].attr4);
bias3 = bias3 - 0.2;
} else {
};*/
//cout << i << "flower" << endl;
//cout << "weight1 is [" << weight1[0] << "," << weight1[1] << ","
//<< weight1[2] << "," << weight1[3] << "]" << endl;
//cout << "weight2 is [" << weight2[0] << "," << weight2[1] << ","
//<< weight2[2] << "," << weight2[3] << "]" << endl;
//cout << "weight3 is [" << weight3[0] << "," << weight3[1] << ","
//<< weight3[2] << "," << weight3[3] << "]" << endl;
//cout << y << "," << y2 << endl;
//cout << "bias:" << bias << "," << bias2 << endl;
} else if (tuples1[i].attr5 == "Iris-versicolor") { //read the lable of trained data
if (y > 0) { //judge and adjust weight1
weight1[0] = weight1[0] - 0.1 * (tuples1[i].attr1);
weight1[1] = weight1[1] - 0.1 * (tuples1[i].attr2);
weight1[2] = weight1[2] - 0.1 * (tuples1[i].attr3);
weight1[3] = weight1[3] - 0.1 * (tuples1[i].attr4);
bias = bias - 0.1;
} else {
}
if (y2 > 0) {
} //judge and adjust weight2
else {
weight2[0] = weight2[0] + 0.1 * (tuples1[i].attr1);
weight2[1] = weight2[1] + 0.1 * (tuples1[i].attr2);
weight2[2] = weight2[2] + 0.1 * (tuples1[i].attr3);
weight2[3] = weight2[3] + 0.1 * (tuples1[i].attr4);
bias2 = bias2 + 0.1;
}
/*if (y3 > 0) { //判断第三个权值是否合适
weight3[0] = weight3[0] - 0.2 * (tuples1[i].attr1);
weight3[1] = weight3[1] - 0.2 * (tuples1[i].attr2);
weight3[2] = weight3[2] - 0.2 * (tuples1[i].attr3);
weight3[3] = weight3[3] - 0.2 * (tuples1[i].attr4);
bias3 = bias3 - 0.2;
} else {
}*/
//cout << i << "flower" << endl;
//cout << "weight1 is [" << weight1[0] << "," << weight1[1] << ","
//<< weight1[2] << "," << weight1[3] << "]" << endl;
//cout << "weight2 is [" << weight2[0] << "," << weight2[1] << ","
//<< weight2[2] << "," << weight2[3] << "]" << endl;
//cout << "weight3 is [" << weight3[0] << "," << weight3[1] << ","
//<< weight3[2] << "," << weight3[3] << "]" << endl;
//cout << y << "," << y2 << endl;
//cout << "bias:" << bias << "," << bias2 << endl;
} else { //realize the class of trained data
if (y > 0) { //judge and adjust weight1
weight1[0] = weight1[0] - 0.1 * (tuples1[i].attr1);
weight1[1] = weight1[1] - 0.1 * (tuples1[i].attr2);
weight1[2] = weight1[2] - 0.1 * (tuples1[i].attr3);
weight1[3] = weight1[3] - 0.1 * (tuples1[i].attr4);
bias = bias - 0.1;
} else {
}
if (y2 > 0) { //judge and adjust weight2
weight2[0] = weight2[0] - 0.1 * (tuples1[i].attr1);
weight2[1] = weight2[1] - 0.1 * (tuples1[i].attr2);
weight2[2] = weight2[2] - 0.1 * (tuples1[i].attr3);
weight2[3] = weight2[3] - 0.1 * (tuples1[i].attr4);
bias2 = bias2 - 0.1;
} else {
}
/*if (y3 > 0) {
} //判断第三个权值是否合适
else {
weight3[0] = weight3[0] + 0.2* (tuples1[i].attr1);
weight3[1] = weight3[1] + 0.2 * (tuples1[i].attr2);
weight3[2] = weight3[2] + 0.2 * (tuples1[i].attr3);
weight3[3] = weight3[3] + 0.2 * (tuples1[i].attr4);
bias3 = bias3 + 0.2;


}*/
//cout << i << "flower" << endl;
//cout << "weight3 is [" << weight3[0] << "," << weight3[1] << ","
//<< weight3[2] << "," << weight3[3] << "]" << endl;
//cout << y << "," << y2 << endl;
}


}
//cout<<"weight1 is ["<<weight1[0]<<","<<weight1[1]<<","<<weight1[2]<<","<<weight1[3]<<"]"<<endl;
//cout<<"weight2 is ["<<weight2[0]<<","<<weight2[1]<<","<<weight2[2]<<","<<weight2[3]<<"]"<<endl;
//cout<<"weight3 is ["<<weight3[0]<<","<<weight3[1]<<","<<weight3[2]<<","<<weight3[3]<<"]"<<endl;
//cout<<"this is the "<<t<<" time training"<<endl;
/*for (vector<Tuple1>::size_type i = 0; i != tuples1.size(); ++i) { //计算MSE决定何时停止训练
//error=0;
y = weight1[0] * tuples1[i].attr1 + weight1[1] * tuples1[i].attr2
+ weight1[2] * tuples1[i].attr3
+ weight1[3] * tuples1[i].attr4 + bias;
y2 = weight2[0] * tuples1[i].attr1 + weight2[1] * tuples1[i].attr2
+ weight2[2] * tuples1[i].attr3
+ weight2[3] * tuples1[i].attr4 + bias2;
y3 = weight3[0] * tuples1[i].attr1 + weight3[1] * tuples1[i].attr2
+ weight3[2] * tuples1[i].attr3
+ weight3[3] * tuples1[i].attr4 + bias3;*/
/*if (tuples1[i].attr5 == "Iris-setosa") {
if (y > 0) {
} else {
error = error + (-y);
}
if (y2 > 0) {
error = error + y2;


}
}
if (tuples1[i].attr5 == "Iris-versicolor") {


if (y2 > 0) {
} else {
error = error + (-y2);
}


} else {
if (y2 > 0) {
error = error + y2;


}
if (y3 > 0) {
} else {
error = error + (-y3);
}
}


}


//cout << "error is " << error << endl;
cout << "this is " << t << "time" << endl;


}*/
cout << "weight1 is [" << weight1[0] << "," << weight1[1] << ","//output weight and bias
<< weight1[2] << "," << weight1[3] << "]" << endl;
cout << "weight2 is [" << weight2[0] << "," << weight2[1] << ","
<< weight2[2] << "," << weight2[3] << "]" << endl;
cout << "bias:" << bias << "," << bias2 << endl;
cout << "this is " << t << " time" << endl;
if (weight1tep[0] == weight1[0] && weight2tep[0] == weight2[0]//if the weight never change, break loop
&& weight1tep[1] == weight1[1] && weight2tep[1] == weight2[1]
&& weight1tep[2] == weight1[2] && weight2tep[2] == weight2[2]
&& weight1tep[3] == weight1[3] && weight2tep[3] == weight2[3]) {
break;
}
}


}

测试代码把权值输入进去了把不要的删了就行了，第一次神经网络作业，弄了两个晚上，女朋友都没陪好，赶紧赎罪去哈哈哈

【机器学习】多层感知机
2401_84149564的博客
05-11 851
感知机是一种二类分类的线性模型，通过分离超平面将实例划分为正负两类。其学习算法基于误分类的损失函数，利用梯度下降法进行优化。感知机模型简单易实现，但仅能处理线性问题。为突破这一限制，多层感知机（MLP）被提出，通过引入隐含层和非线性激活函数，增强了模型的表达能力。MLP由输入层、隐含层和输出层组成，通过反向传播算法进行训练，广泛应用于图像识别、语音识别等领域。反向传播算法通过前向传播计算输出，反向传播误差并更新权重，以最小化损失函数。本文还提供了使用Python实现MLP的代码示例，包括手动实现和使用PyT
【机器学习】- 感知机（mnist数据集）
cxq_baby的博客
07-29 2254
算法： 感知机算法计算的超平面S会因为权值的初始值和误分类点的选择顺序变化 性能 在对946个测试样本进行测试之后发现，错误率仅为0.42% 数据集和代码可以在这里下载 CODE from os import listdir import numpy as np #读取数据 def readData(fileFolder): fileList = listdir(fileFolder)...
Iris Classification 鸢尾花分类
最新发布
qq_26817675的博客
04-29 692
根据鸢尾花的四个特征萼片长度（ Sepal length）, 萼片宽度（Sepal width）, 花瓣长度（Petal length）, 花瓣宽度（Petal width），区分鸢尾花的三个品种山鸢尾（Iris-setosa）、变色鸢尾（Iris-versicolor）和维吉尼亚鸢尾（Iris-virginica）。实现基础的三分类问题
感知机分类数据(线性可分和不可分)
12-16
感知机分类数据，包括线性可分数据（15.dat）和线性不可分数据（18.dat）。数据来源为林轩田老师的机器学习基石的作业。
数据挖掘C++实现(iris数据集)
01-18
cpp文件，给予iris数据集，cure方法实现
knn分类器c++实现（含测试数据IRIS）
06-19
在vs2008上用c++实现的knn分类器，内含测试数据，是初学者的好东西
c++ iris-贝叶斯分类
09-25
关于iris 的贝叶斯分类，用C++实现的
利用感知机实现鸢尾花分类问题
婷子的博客
12-16 5486
1. 实验简介 感知机是一种二分类的线性判别模型。 2. 实验原理 3. 实验代码 3.1 方法一 3.2 方法二
【数据挖掘】KNN算法详解及对iris数据集分类实战（超详细 附源码）
showswoller的博客
01-08 1050
【数据挖掘】KNN算法详解及对iris数据集分类实战（超详细 附源码）
鸢尾花SVM分类数据集：MATLAB与Python实现指南
鸢尾花数据集（Iris dataset）是机器学习和统计分类中常用的一个数据集，最早由英国生物学家、统计学家罗纳德·费舍尔（Ronald Fisher）于1936年在论文中提出，用于展示线性判别分析的效果。该数据集包含了150个样本...
Opencv学习记录【十】——神经网络MLP(多层感知机)
stupid_miao的博客
05-09 2575
opencv为我们提供了多种机器学习方法，比如adaboost、svm、神经网络等。本文主要记录其神经网络的原理和用法(参考赵春江的“机器学习经典算法剖析基于ｏｐｅｎｃｖ”，人民邮电出版社，214-220) 下面就以照片的方式贴出其原理： 如何利用ｏｐｅｎｃｖ的神经网络ａｐｉ呢，很简单，可以看其官网或者按照下面的例子进行实践。 这个例子是人脸贴图利用ｏｐｅｎｃｖ神经网络对脸部区域的位置和脸部器官...
运用感知机算法的线性分类程序
01-10
感知机算法，采用感知机算法设计分类器，并用其中的数据来验证测试分类器性能
感知机算法实现(使用MNIST数据集)_Python环境
05-16
在Python环境下实现感知机算法(使用MNIST数据集)，代码有详细注释，使用的是感知机算法的原始形式
对iris数据进行分组分类
03-15
对IRIS数据集进行分组，分类
minist数据集 csv格式（适用于本人感知机博文）
01-16
minist数据集的csv格式 网上的其他资源均为gz等格式，所以找了一份csv格式的数据上传分享给大家
iris数据集分类
07-17
简单但是实用的程序，适合进行数值计算，其中包含线性和bp网络的分析。
感知器算法
cuixuange的博客
10-18 4869
**感知器算法: 线性判别函数 或者说广义的线性判别曲面 进行二分类或者多分类的代码实现 以及使用iris_data 的感知机测试代码 不断地迭代赏罚 为后面的梯度下降方式做铺垫** -github：https://github.com/cuixuage/Machine_Learning 参考资料 https://ljalphabeta.gitbooks.io/python-/conte
机器学习入门实战——感知机算法实战Iris数据集
热门推荐
Quanfita的博客
02-25 1万+
感知机算法实战Iris数据集 关于感知机的相关理论知识请查看：感知机 关于Iris数据集 Iris也称鸢尾花卉数据集，是一类多重变量分析的数据集。数据集包含150个数据集，分为3类，每类50个数据，每个数据包含4个属性。可通过花萼长度，花萼宽度，花瓣长度，花瓣宽度4个属性预测鸢尾花卉属于（Setosa，Versicolour，Virginica）三个种类中的哪一类。 Iris...
感知机
watermelon12138的博客
03-31 1004
一、概念 感知机是用于二分类的线性分类模型，其输入是实例的特征向量，输出是实例的类别，类别取+1和-1二个值，+1代表正类，-1代表负类。感知机对应于输入空间（特征空间）中将实例分为正负两类的分割超平面，属于判别模型。感知机学习算法简单易于实现，分为原始形式和对偶形式。 生成模型和判别模型 监督学习方法可以分为生成方法和判别方法，对应所学到的模型称为生成模型和判别模型。 生成模型：生成方法通过数据...
评论
添加红包

请填写红包祝福语或标题

个

红包个数最小为10个

元

红包金额最低5元

当前余额3.43元 前往充值 >
需支付：10.00元
成就一亿技术人!
领取后你会自动成为博主和红包主的粉丝 规则
hope_wisdom
发出的红包
实付元
使用余额支付
点击重新获取
扫码支付
钱包余额 0

抵扣说明：

1.余额是钱包充值的虚拟货币，按照1:1的比例进行支付金额的抵扣。
2.余额无法直接购买下载，可以购买VIP、付费专栏及课程。

余额充值