深度学习之Pytorch实现房价预测

• 废话不多说
• 准备数据

# 为实验准备数据
def init_data():
   x = torch.FloatTensor(torch.linspace(0, 100))
   y = x + torch.FloatTensor(torch.randn(100)) * 10
   return x, y

def train_test_split(x, y, radio=0.1):
	assert len(x) == len(y), "length of x is not equal to y"
	size = int(len(x) * radio)
	# print(size)
	# print(x.shape)
	x_train, x_test = x[:(len(x) - size)], x[(len(x) - size):]
	y_train, y_test = y[:(len(y) - size)], y[(len(y) - size):]
	return x_train, x_test, y_train, y_test

训练数据可视化

def plot(x_train,  y_train):
	plt.figure(figsize=(10,8))
	plt.plot(x_train.data.numpy(),y_train.data.numpy(),'o')
	plt.xlabel("X")
	plt.ylabel("Y")
	plt.show()

• 来来来你想要的梯度下降法

x, y = init_data()
	print(x.shape)
	x_train, x_test, y_train, y_test = train_test_split(x, y)
	plot(x_train,y_train)
	w = torch.nn.Parameter(torch.randn(1))
	b = torch.nn.Parameter(torch.randn(1))
	learn_rate = 0.0001
	# print(len(x_train))
	# print(len(x_test))
	for i in range(1000):
		pred = w.expand_as(x_train)*x_train+b.expand_as(x_train)
		loss  = torch.mean((pred-y_train)**2)
		print(loss)
		loss.backward()
		w.data.add_(-learn_rate*w.grad.data)
		b.data.add_(-learn_rate*b.grad.data)
		w.grad.data.zero_()
		b.grad.data.zero_()

	x_data = x_train.data.numpy()
	plt.figure(figsize=(10,7))
	xplot, = plt.plot(x_data,y_train.data.numpy(),'o')
	yplot, = plt.plot(x_data,w.data.numpy()*x_data+b.data.numpy())
	plt.xlabel("X")
	plt.ylabel("Y")
	str1 = str(w.data.numpy()[0])+'x+'+str(b.data.numpy()[0])
	plt.legend([xplot,yplot],['Data',str1])
	plt.show()
	pred = w.expand_as(x_test)*x_test+b.expand_as(x_test)
	print(pred)

	x_data = x_train.data.numpy()
	x_pred = x_test.data.numpy()

	plt.figure(figsize=(10,7))
	plt.plot(x_data,y_train.data.numpy(),'o')
	plt.plot(x_pred,y_test.data.numpy(),'s')
	x_data = np.r_[x_data,x_test.data.numpy()]
	plt.plot(x_data,w.data.numpy()*x_data+b.data.numpy())
	plt.plot(x_pred,w.data.numpy()*x_pred+b.data.numpy(),'o')
	plt.xlabel("X")
	plt.ylabel("Y")
	str1 = str(w.data.numpy()[0]) + 'x+' + str(b.data.numpy()[0])
	plt.legend([xplot, yplot], ['Data', str1])
	plt.show()

• 训练集结果显示
  
• 训练集待判结果显示
• 
• 待判和回判结果显示
• 
• 理解pytorch底层原理
• 梯度下降法的的基本实现
• matplotlib基本绘图