Gradient descent optimization in ML Python - Interactive Code Practice

Practice - 5 Tasks

1fill in blank

easy

Complete the code to update the weight using gradient descent.

ML Python

weight = weight - [1] * gradient

Drag options to blanks, or click blank then click option'

Aepoch

Bmomentum

Clearning_rate

Dloss

Attempts:

3 left

💡 Hint

Common Mistakes

Using 'momentum' instead of 'learning_rate' for the step size.

Using 'loss' which is a measure, not a step size.

Using 'epoch' which counts iterations, not step size.

The learning rate controls how big a step we take in the direction of the negative gradient to update the weight.

2fill in blank

medium

Complete the code to calculate the gradient of the loss with respect to the weight.

ML Python

gradient = 2 * (prediction - target) * [1]

Drag options to blanks, or click blank then click option'

Alearning_rate

Bweight

Cbias

Depoch

Attempts:

3 left

💡 Hint

Common Mistakes

Using 'learning_rate' which is unrelated to gradient calculation.

Using 'bias' which is a separate parameter.

Using 'epoch' which counts iterations.

The gradient is calculated by multiplying the error by the derivative of the prediction with respect to the weight, which is the input feature or weight itself depending on the model.

3fill in blank

hard

Fix the error in the code to correctly perform one step of gradient descent.

ML Python

weight = weight [1] learning_rate * gradient

Drag options to blanks, or click blank then click option'

A-

B+

C*

D/

Attempts:

3 left

💡 Hint

Common Mistakes

Using '+' which moves weights in the wrong direction.

Using '*' or '/' which are not correct operators for updating weights.

Gradient descent updates weights by moving in the negative direction of the gradient, so we subtract the product of learning rate and gradient.

4fill in blank

hard

Fill both blanks to create a dictionary of squared errors for each data point where error is less than 1.

ML Python

squared_errors = {x: (y - prediction)[1]2 for x, y in data.items() if abs(y - prediction) [2] 1}

Drag options to blanks, or click blank then click option'

A**

B<

C>

D-

Attempts:

3 left

💡 Hint

Common Mistakes

Using '+' instead of '**' for squaring.

Using '>' instead of '<' for filtering.

We use '**' to square the error and '<' to filter errors less than 1.

5fill in blank

hard

Fill all three blanks to create a dictionary of weights updated by gradient descent for each feature.

ML Python

updated_weights = {feature: weights[feature] [1] learning_rate * gradients[[2]] for feature in features if gradients[feature] [3] 0}

Drag options to blanks, or click blank then click option'

A-

Bfeature

C>

D+

Attempts:

3 left

💡 Hint

Common Mistakes

Using '+' instead of '-' for weight update.

Using '+' or '<' instead of '>' for filtering gradients.

Weights are updated by subtracting learning rate times gradient. The gradient is accessed by the feature key. We filter only positive gradients using '>'.