1.1 鸢尾花数据集介绍

iris数据集是用来给莺尾花做分类的数据集，每个样本包含了花萼长度、花萼宽度、花瓣长度、花瓣宽度四个特征，我们需要建立一个分类器，该分类器可通过样本的四个特征来来判断样本属于山鸢尾（Setosa）、变色鸢尾（Versicolour）还是维吉尼亚鸢尾（Virginica）中的哪一个，选择神经网络进行分类。

1.2 思路流程

• 导入鸢尾花数据集
• 对数据集进行切分，分为训练集和测试集
• 搭建网络模型
• 训练网络
• 将所训练出的模型进行保存（准确率大于90%）

1.3 网络模型

采用sigmoid等函数，算激活函数时（指数运算），计算量大，反向传播求误差梯度时，求导涉及除法，计算量相对大，而采用Relu激活函数，整个过程的计算量节省很多，故采用Relu作为激活函数

1.4 实现代码

导入所需要的的模块

import torch
import torch.nn as nn
from sklearn import datasets
from sklearn.model_selection import train_test_split

神经网络类

class Net(nn.Module):def __init__(self,in_num,out_num,hid_num):super(Net,self).__init__()self.network = nn.Sequential(nn.Linear(in_num,hid_num),nn.ReLU(),nn.Linear(hid_num,out_num))self.optimizer = torch.optim.SGD(self.parameters(), lr=0.05)self.loss_func = torch.nn.CrossEntropyLoss()def forward(self,x):return self.network(x)def train(self,x,y):out = self.forward(x)loss = self.loss_func(out,y)self.optimizer.zero_grad()loss.backward()self.optimizer.step()print('loss = %.4f' % loss.item())def test(self,x):return self.forward()

引入数据集，并按照8:2切分训练集和测试集

dataset = datasets.load_iris()
input = torch.FloatTensor(dataset['data'])
label = torch.LongTensor(dataset['target'])
x_train, x_test, y_train, y_test = train_test_split(input, label, test_size=0.2)

如果存在已有训练好的网络则导入，并在总体数据集上测试其准确性

try:print("iris_model exist and have been loaded")mynet = torch.load('iris_model.pkl')output = mynet(input)pred_y = torch.max(output, 1)[1].numpy()sum = 0for i in range(len(label)):if pred_y[i] == label[i]:sum = sum + 1accuracy = float(sum / len(label))print('model accuracy = %d%% (testing on the whole dataset)' % (accuracy * 100))

若不存在训练好的网络则进行训练，直到准确性大于90%后将其保存

except:mynet = Net(4,10,3)accuracy = 0.0while accuracy < 0.9:for i in range (10000):mynet.train(x_train,y_train)output = mynet(x_test)pred_y = torch.max(output, 1)[1].numpy()sum=0for i in range(len(y_test)):if pred_y[i] == y_test[i]:sum=sum+1accuracy = float(sum / len(y_test))torch.save(mynet, 'iris_model.pkl')print(mynet)print("The net have been saved")print('accuracy = %d%%' % (accuracy*100))

鸢尾花识别完整代码

import torch
import torch.nn as nn
from sklearn import datasets
from sklearn.model_selection import train_test_split
class Net(nn.Module):def __init__(self,in_num,out_num,hid_num):super(Net,self).__init__()self.network = nn.Sequential(nn.Linear(in_num,hid_num),nn.ReLU(),nn.Linear(hid_num,out_num))self.optimizer = torch.optim.SGD(self.parameters(), lr=0.05)self.loss_func = torch.nn.CrossEntropyLoss()def forward(self,x):return self.network(x)def train(self,x,y):out = self.forward(x)loss = self.loss_func(out,y)self.optimizer.zero_grad()loss.backward()self.optimizer.step()print('loss = %.4f' % loss.item())def test(self,x):return self.forward()if __name__ == '__main__':dataset = datasets.load_iris()input = torch.FloatTensor(dataset['data'])label = torch.LongTensor(dataset['target'])x_train, x_test, y_train, y_test = train_test_split(input, label, test_size=0.2)try:print("iris_model exist and have been loaded")mynet = torch.load('iris_model.pkl')output = mynet(input)pred_y = torch.max(output, 1)[1].numpy()sum = 0for i in range(len(label)):if pred_y[i] == label[i]:sum = sum + 1accuracy = float(sum / len(label))print('model accuracy = %d%% (testing on the whole dataset)' % (accuracy * 100))except:mynet = Net(4,10,3)accuracy = 0.0while accuracy < 0.9:for i in range (10000):mynet.train(x_train,y_train)output = mynet(x_test)pred_y = torch.max(output, 1)[1].numpy()sum=0for i in range(len(y_test)):if pred_y[i] == y_test[i]:sum=sum+1accuracy = float(sum / len(y_test))torch.save(mynet, 'iris_model.pkl')print(mynet)print("The net have been saved")print('accuracy = %d%%' % (accuracy*100))

github文件链接