神经网络防止过拟合

Dropout Regularization

通过带有正则化的神经网络和没有正则化的神经网路作对比，从直观上观察正则化的效果，附上代码：

import torch
import matplotlib.pyplot as plt

N_SAMPLES = 20 #每种数据的总数
N_HIDDEN = 300 #网络中的节点数

# 训练数据
x = torch.unsqueeze(torch.linspace(-1, 1, N_SAMPLES), 1)
y = x + 0.3*torch.normal(torch.zeros(N_SAMPLES, 1), torch.ones(N_SAMPLES, 1))

# 测试数据
test_x = torch.unsqueeze(torch.linspace(-1, 1, N_SAMPLES), 1)
test_y = test_x + 0.3*torch.normal(torch.zeros(N_SAMPLES, 1), torch.ones(N_SAMPLES, 1))

# 展示训练和测试数据
# plt.scatter(x.data.numpy(), y.data.numpy(), c='magenta', s=50, alpha=0.5, label='train')
# plt.scatter(test_x.data.numpy(), test_y.data.numpy(), c='cyan', s=50, alpha=0.5, label='test')
# plt.legend(loc='upper left')
# plt.ylim((-2.5, 2.5))
# plt.show()

#构造产生过拟合的神经网络
net_overfitting = torch.nn.Sequential(
    torch.nn.Linear(1, N_HIDDEN),
    torch.nn.ReLU(),
    torch.nn.Linear(N_HIDDEN, N_HIDDEN),
    torch.nn.ReLU(),
    torch.nn.Linear(N_HIDDEN, 1),
)

#构造带有Dropout的神经网络
net_dropped = torch.nn.Sequential(
    torch.nn.Linear(1, N_HIDDEN),
    torch.nn.Dropout(0.5),  # 减少50%的神经元
    torch.nn.ReLU(),
    torch.nn.Linear(N_HIDDEN, N_HIDDEN),
    torch.nn.Dropout(0.5),
    torch.nn.ReLU(),
    torch.nn.Linear(N_HIDDEN, 1),
)

print(net_overfitting)  # 输出网络体系结构
print(net_dropped)
#优化器optimizer
optimizer_ofit = torch.optim.Adam(net_overfitting.parameters(), lr=0.01)
optimizer_drop = torch.optim.Adam(net_dropped.parameters(), lr=0.01)
loss_func = torch.nn.MSELoss()

plt.ion()

for t in range(500):
    pred_ofit = net_overfitting(x)     #向网络里传数据，生成结果
    pred_drop = net_dropped(x)
    loss_ofit = loss_func(pred_ofit, y) #计算误差
    loss_drop = loss_func(pred_drop, y)

    optimizer_ofit.zero_grad()  #上一步的更新梯度留在网络参数中，保存在grad里
    optimizer_drop.zero_grad()
    loss_ofit.backward()   # 误差反向传播, 计算参数更新值
    loss_drop.backward()
    optimizer_ofit.step()  # 更新网络中的参数
    optimizer_drop.step()

    if t % 10 == 0:
        # 更改为预测模式，没有dropout效果
        net_overfitting.eval()
        net_dropped.eval()

        plt.cla()
        test_pred_ofit = net_overfitting(test_x)
        test_pred_drop = net_dropped(test_x)

        #可视化
        plt.scatter(x.data.numpy(), y.data.numpy(), c='magenta', s=50, alpha=0.3, label='train')
        plt.scatter(test_x.data.numpy(), test_y.data.numpy(), c='cyan', s=50, alpha=0.3, label='test')
        plt.plot(test_x.data.numpy(), test_pred_ofit.data.numpy(), 'r-', lw=3, label='overfitting')
        plt.plot(test_x.data.numpy(), test_pred_drop.data.numpy(), 'b--', lw=3, label='dropout(50%)')
        plt.text(0, -1.2, 'overfitting loss=%.4f' % loss_func(test_pred_ofit, test_y).data.numpy(), fontdict={'size': 20, 'color':  'red'})
        plt.text(0, -1.5, 'dropout loss=%.4f' % loss_func(test_pred_drop, test_y).data.numpy(), fontdict={'size': 20, 'color': 'blue'})
        plt.legend(loc='upper left'); plt.ylim((-2.5, 2.5));plt.pause(0.1)

        # 更改为测试模式，有dropout效果
        net_overfitting.train()
        net_dropped.train()

plt.ioff()
plt.show()

输出最终结果

注意

• 在训练网络时，通过.train()使得网络带有Dropout效果
• 在测试数据时，通过.eval()使得网络不带有Dropout效果