def compute_acc(X, Y, W):
y_hat = sr.compute_a(sr.compute_z(X, W, 0))
y_hat_indicies = np.argmax(y_hat, axis=1)
y_indicies = np.argmax(Y, axis=1)
error = np.array([y_indicies == y_hat_indicies])
acc = error.sum() / X.shape[0]
return acc
def compute_z2(a1, W2, b2):
'''
Compute the linear logit values of a data instance in the first layer. z1 = W1 x + b1
Input:
a1: the non-linear activations in the first layer, a float numpy vector of shape h by 1.
W2: the weight matrix of the 2nd layer, a float numpy matrix of shape (h by c). Here c is the number of classes.
b2: the bias values of the 2nd layer, a float numpy vector of shape c by 1.
Output:
z2: the linear logits of the 2nd layer, a float numpy vector of shape c by 1.
'''
z2 = sr.compute_z(a1, W2, b2)
return z2
def compute_z1(x, W1, b1):
'''
Compute the linear logit values of a data instance in the first layer. z1 = W1 x + b1
Input:
x: the feature vector of a data instance, a float numpy vector of shape p by 1. Here p is the number of features/dimensions.
W1: the weight matrix of the first layer, a float numpy matrix of shape (h by p). Here h is the number of outputs in the first layer.
b1: the bias values of the first layer, a float numpy vector of shape h by 1.
Output:
z1: the linear logits, a float numpy vector of shape h by 1.
Hint: you could reuse the fucntions in problem 1, for example sr.function_name()
'''
z1 = sr.compute_z(x, W1, b1)
return z1