The learning problem - assignment

The learning problem - assignment

All the questions below refer to the the cells of the notebook numbered 1.0 and 1.1 (aka not 1.26).

Github (10 points)

Clone the repo in your own Github account. Make the Github repo private. Locate the notebook shown above.

Run the curve fitting notebook (10 points)

For the notebook embedded in {{}}, load it in your Colab account and ensure it runs with no errors.

Loss function (15 points)

Add a new text cell and explain why the loss function is called Root Mean Squared Error (RMSE) and what the square root offers, if anything, to finding a better solution.

Model complexity (15 points)

Add a text cell after the corresponding figure explaining the behavior of the test error for M=9 vs M=3.

Parameter Table (20 points)

Extract the weights \(w_i\) (parameters) of the model for M=1, 3, 6, 9 and tabulate them into corresponding columns. What can you observe with respect to their squared norm \(|\mathbf w|\) ?

Regularization (20 points)

Read the Linear Regression notes in the course site and add a text paragraph after the regularization cell explaining the output. Create a plot of the RMSE vs model complexity with regularization.

Github (10 points)

Ensure that you save the notebook with the results / plots and commit the notebook to your Github account. Submit (a) the URL of the Github hosted notebook and (b) the URL of the Colab histed URL in the corresponding Assignment in Canvas.