How to perform Numpy optimisation for this code?

Question

I have the following code snippet:

def func1(self, X, y):
    #X.shape = (455,13)
    #y.shape = (455)

    num_examples, num_features = np.shape(X)
    self.weights = np.random.uniform(-1 / (2 * num_examples), 1 / (2 * num_examples), num_features)

    while condition:
        new_weights = np.zeros(num_features)
        K = (np.dot(X, self.weights) - y)

        for j in range(num_features):
            summ = 0

            for i in range(num_examples):
                summ += K[i] * X[i][j]

            new_weights[j] = self.weights[j] - ((self.alpha / num_examples) * summ)

        self.weights = new_weights

This code works too slow. Is there any optimization, which I can do?

Do you really want to reset summ=0 inside the for j loop? By doing so, you throw away all the work done by every iteration of the for j loop except the last iteration when j equals num_features-1. — unutbu
– unutbu, Commented May 3, 2015 at 14:30

Saullo G. P. Castro · Accepted Answer · 2015-05-03 14:44:05Z

4

You can efficiently use np.einsum(). See a testing version below:

def func2(X, y):
    num_examples, num_features = np.shape(X)
    weights = np.random.uniform(-1./(2*num_examples), 1./(2*num_examples), num_features)

    K = (np.dot(X, weights) - y)

    return weights - alpha/num_examples*np.einsum('i,ij->j', K, X)

answered May 3, 2015 at 14:44

Saullo G. P. Castro

59.5k28 gold badges191 silver badges244 bronze badges

Sign up to request clarification or add additional context in comments.

Comments

Divakar · Accepted Answer · 2015-05-03 14:43:20Z

2

You can get new_weights directly using matrix-multiplication with np.dot like so -

new_weights = self.weights- ((self.alpha / num_examples) * np.dot(K[None],X))

answered May 3, 2015 at 14:43

Divakar

222k19 gold badges273 silver badges374 bronze badges

2 Comments

Denis Over a year ago

thanks for the help! It seems that your code works correctly, but a bit slower than Saullo Castro.

Divakar Over a year ago

@Denis Yeah, that einsum seems to be the best one for optimizing solutions!

Collectives™ on Stack Overflow

How to perform Numpy optimisation for this code?

2 Answers 2

Comments

2 Comments

Your Answer

Hot Network Questions

Collectives™ on Stack Overflow

2 Answers 2

Comments

2 Comments

Your Answer

Sign up or log in

Post as a guest

Related