return probs,preds
def getAccuracy(someX, someY, w):
prob,prede = getProbsAndPreds(someX, w)
accuracy = sum(prede == someY)/(float(len(someY)))
return accuracy
def predict(x, w):
out = np.array([np.argmax(x) for x in softmax(x.dot(w))])
return out
def score(x, y):
prediction = predict(x)
acc = 1 / len(y) * sum([(prediction[i] == y[i]) for i in range(len(y))])
return acc
if __name__ == "__main__":
data, label = read_data.read_data_from_resource()
# minmax正则化：否则数据不好收敛，到nan
min1 = np.min(data[:, 1])
max1 = np.max(data[:, 1])
min2 = np.min(data[:, 2])
max2 = np.max(data[:, 2])
m, n = np.shape(data)
for i in range(m):
data[i][1] = (data[i][1] - min1) / (max1 - min1)
data[i][2] = (data[i][2] - min2) / (max2 - min2)
w, losses, thetaList = softmaxRegressionSGD(data, label, 5000)
print('Training Accuracy: ', getAccuracy(data, label, w))
# plt.plot(losses)
# plt.legend() # 将样例显示出来
# plt.show()
error = classLabels[random] - h
new_weights = weights + alpha * error * dataMatIn[random]
weights = new_weights
# for i in range(m):
# h = sigmoid(np.sum(new_weights * dataMatIn[i]))
# loss = loss_funtion(dataMatIn, classLabels, weights)
h = sigmoid(np.sum(new_weights * dataMatIn[random]))
loss = loss_funtion(dataMatIn, classLabels, weights)
loss_array.append(loss)
theta_array.append(weights)
iter_count += 1
return weights, loss_array, theta_array
data, labels = read_data_from_resource("dataset1")
# minmax正则化：否则数据不好收敛，到nan
min1 = np.min(data[:, 1])
max1 = np.max(data[:, 1])
min2 = np.min(data[:, 2])
max2 = np.max(data[:, 2])
m, n = np.shape(data)
for i in range(m):
data[i][1] = (data[i][1] - min1) / (max1 - min1)
data[i][2] = (data[i][2] - min2) / (max2 - min2)
max_iters = 10000
weight, loss_array, theta_array = stocGradDescent(data, labels, max_iters)
print("theta is :", theta_array[-1])
print("loss", loss_array[-1])
cost_array = []
weight_array = []
for epoch in range(epochs):
z = np.dot(X, weight)
activation = softmax(z)
diff = activation - y
grad = np.dot(X.T, diff)
weight -= learning_rate * grad
cost = compute_cost(X, y, weight)
cost_array.append(cost)
weight_array.append(np.array(weight).squeeze())
return weight, cost_array, weight_array
# 这里修改了: data.insert(0,1.0)
data, label = read_data_from_resource("dataset2")
# minmax正则化：否则数据不好收敛，到nan
min1 = np.min(data[:, 1])
max1 = np.max(data[:, 1])
min2 = np.min(data[:, 2])
max2 = np.max(data[:, 2])
m, n = np.shape(data)
for i in range(m):
data[i][1] = (data[i][1] - min1) / (max1 - min1)
data[i][2] = (data[i][2] - min2) / (max2 - min2)
labels = one_hot_encoder(label, m)
learned_w, cost_array, weight_array = softmax_gradient_descent(data, labels)
print("Learned weights:")
print(learned_w)
else:
xcord2.append(dataArr[i, 1])
ycord2.append(dataArr[i, 2])
fig = plt.figure()
ax = fig.add_subplot(111)
ax.scatter(xcord1, ycord1, s=30, c='red', marker='s')
ax.scatter(xcord2, ycord2, s=30, c='green')
x = np.arange(-0.1, 1.0, 0.01)
y = (-weights[0] - weights[1] * x) / weights[2]
ax.plot(x, y)
plt.xlabel('X')
plt.ylabel('Y')
plt.show()
dataMat, labelMat = read_data_from_resource()
# minmax正则化：否则数据不好收敛，到nan
min1 = np.min(dataMat[:, 1])
max1 = np.max(dataMat[:, 1])
min2 = np.min(dataMat[:, 2])
max2 = np.max(dataMat[:, 2])
m, n = np.shape(dataMat)
for i in range(m):
dataMat[i][1] = (dataMat[i][1] - min1) / (max1 - min1)
dataMat[i][2] = (dataMat[i][2] - min2) / (max2 - min2)
dataArr = np.array(dataMat)
weights, losst, loss_array, theta_array = gradDescent(dataArr, labelMat)
print("weight:", weights)
print("loss:", losst)
# 数据可视化
plotBestFit(dataArr, labelMat, weights)