# -*- coding: gbk -*- import torch import pandas as pd import torch.nn as nn import torch.utils.data as Data import torchvision # 数据库模块 import matplotlib.pyplot as plt import torch.nn.functional as F # 激励函数都在这 import numpy as np import csv LR=0.01 BATCH_SIZE=25 tem=pd.read_csv("wine.csv",sep=',') tem=np.array(tem.iloc[:,:]).reshape(177,14) np.random.shuffle(tem) Row=177 Half=80 Col=14 x0_train=tem[0:Half,1:Col].reshape(Half,13) y0_train=tem[0:Half,0] x0_test=tem[Half:Row,1:Col].reshape(Row-Half,13) y0_test=tem[Half:Row,0] x=torch.FloatTensor(x0_train) xx=torch.FloatTensor(x0_test) y=torch.FloatTensor(y0_train) yy=torch.FloatTensor(y0_test) y=y.long() yy=yy.long() y=y-1 yy=yy-1 net = torch.nn.Sequential( torch.nn.Linear(13, 100), torch.nn.ReLU(), torch.nn.Linear(100, 3), ) optimizer= torch.optim.Adam(net.parameters(), lr=LR, betas=(0.9, 0.99),weight_decay=1e-5) loss_func = torch.nn.CrossEntropyLoss() torch_dataset = Data.TensorDataset(x,y) loader = Data.DataLoader( dataset=torch_dataset, # torch TensorDataset format batch_size=BATCH_SIZE, # mini batch size shuffle=True, # 要不要打乱数据 (打乱比较好) ) for t in range(3000): cnt=0 for step, (batch_x, batch_y) in enumerate(loader): # 每一步 loader 释放一小批数据用来学习 out = net(batch_x) # 喂给 net 训练数据 x, 输出预测值 loss = loss_func(out, batch_y) # 计算两者的误差 optimizer.zero_grad() # 清空上一步的残余更新参数值 loss.backward() # 误差反向传播, 计算参数更新值 optimizer.step() # 将参数更新值施加到 net 的 parameters 上 if(t%250==0): ac_sum = 0 out = net(xx) for i in range(Row-Half): goal = yy[i].item() if out[i][0] > out[i][1] and out[i][0] > out[i][2]: if (goal == 0): ac_sum += 1 elif out[i][1] > out[i][0] and out[i][1] > out[i][2]: if (goal == 1): ac_sum += 1 else: if (goal == 2): ac_sum += 1 print("轮数： ",t," ","正确率： ", ac_sum, " ", ac_sum / float(Row-Half)) torch.save(net,'wine_net.pkl')
