Python中的逻辑回归

Question

I'm trying to build a Logistic regression model using numPy and training it on TensorFlow "Getting Started" example: {x: [1, 2, 3, 4], y: [0, -1, -2, -3]} using the same learning rate and epochs as the one on tensorFlow example but for some reason it cant learn the correct weight and bias. Any help? I'm new to AI.

Code:

# Compute cost and gradient
def propagate(w, b, X, Y):
    m = X.shape[0]
    A = sigmoid(np.multiply(w,X) + b)
    arr = (np.multiply(w,X) + b) - Y
    cost = np.dot(arr, arr)
    cost = np.squeeze(cost)
    dw = 1/m * X.dot((A-Y).T)
    db = 1/m * np.sum(A-Y)

    return {"db": db, "dw": dw}, cost

# Gradient Descnet
def optimize(w, b, X, Y, epochs, learning_rate):
    costs = []

    for i in range(epochs):
        grads, cost = propagate(w, b, X, Y)
        dw = grads['dw']
        db = grads['db']
        w = w - learning_rate * dw
        b = b - learning_rate * db
        if i % 100 == 0:
            costs.append(cost)

    return {"w":w, "b":b}, {"db": db, "dw": dw}, costs

Output:

w, b, X, Y = np.array([0.3]), -0.3, np.array([1, 2, 3, 4]), np.array([0, -1, -2, -3])

grads, cost = propagate(w, b, X, Y)

print ("dw = " + str(grads["dw"])) # dw = 6.6074129907
print ("db = " + str(grads["db"])) # db = 2.10776208142
print ("cost = " + str(cost))      # cost = 23.66

params, grads, costs = optimize(w, b, X, Y, epochs= 100, learning_rate = 0.01)

print ("w = " + str(params["w"])) # w = [-4.85038348] (supposed to be about -0.9999969)
print ("b = " + str(params["b"])) # b = -1.86763966366 (supposed to be about 0.99999082)

Answer 1

You can just use a linear function as activation for your last layer. And it is better to normalize your data to have better results.