Download as pdf or txt
Download as pdf or txt
You are on page 1of 4

Assignment

M. S. Ramaiah University of Applied Sciences


Faculty of Engineering & Technology

Department Computer Science & Engineering


1. M. Tech. in Artificial Intelligence and Machine Learning
Programme 2. M. Tech. in Data Science and Engineering

Batch 2021-2023
Course Name Artificial Neural Networks
Course Code 19MIC504A
Course Faculty Dr. S. Thompson

Problem Max marks


A.1. 10
A.2. 10
A.3. 10
A.4. 10
A.5. 5
A.6. 5

Submission Information
Submission Date 10-11-2022
Important note:
1. The assignment has to be neatly word processed. Fix font size to 10 and font style
to Times New Roman. Check for formatting issues.
2. The maximum number of pages should not exceed 30.
3. Kindly be aware that the submission after the due date is not permitted and will
reduce marks.
Questions:

1. Would you use regression or classification to address each of the following machine
learning problems? Describe your responses. When in doubt, choose one and state why
you chose it.

a. A website that estimates how much money a person will spend there.

b. A voice assistant translating spoken words into text.

c. Buying or selling shares from a certain business.

d. YouTube suggesting a viewer watch a certain video.

2. Consider the below dataset that depicts the size and price of houses.

Consider that we have developed a model with size-based house price prediction as the
following:

a. Determine the predictions this model makes based on the dataset.

b. Determine this model's mean absolute error.

c. Determine this model's root mean square error.

3. Using the same dataset mentioned in A.2, and four different hyperparameters, we
trained four models. The training and testing errors for each model are listed in the
table below.
a. For this dataset, which model would you choose?

b. Which model appears to be underfitting the data?

c. Which model appears to be too closely fitting the data?

4. Consider the below dataset:

A polynomial regression model is trained for the above dataset and it predicts the value
of y as below:

Determine the following if the error function we used to train this dataset is the mean
absolute value (MAE) and the regularisation parameter is λ= 0.1:

a. The lasso regression error of our model (using the L1 norm)

b. The ridge regression error of our model (using the L2 norm)

5. To determine whether a patient has a decaying tooth, a dentist trained a logistic


classifier on a dataset of patients. According to the model, the chance that a patient has
a decaying tooth is given by the below model:
Where,

• d is a variable that indicates whether the patient has had another decayed tooth in the
past, and

• c is a variable that indicates whether the patient eats candy.

For example, if a patient eats candy, then c = 1, and if they don’t, then c = 0. What is the
probability that a patient that eats candy and was treated for a decayed tooth last year
has a decayed tooth today?

6. The neural network in the below figure only has sigmoid functions as activations. For
the input (1,1), what would this neural network predict?

You might also like