USI Machine Learning Projects

The projects are from the Machine Learning course 2019-2020 of USI, Lugano.
Both Projects were mainly developed using the Keras library, and some scikit-learn.

Project 1

The first Project is about solving a regression problem for a given dataset.
The project is composed of 2 Tasks:

1. Use the family of models f(x, theta) = theta_0 + theta_1 * x_1 + theta_2 * x_2 + theta_3 * sin(x_1) * x_2 to fit the data.
2. Consider any family of non-linear models of your choice to address the regression problem, and compare them with the model from Task 1.

The code for each individual Task can be found in the src directory. To run the script through a terminal, just type the command

$ python3 file1.py

or

$ python3 file2.py

Each script creates a model and saves it in the deliverable directory. There, the command

$ python3 run_model.py

can be run to just see the accuracies and errors of each model when evaluated to the test sets.

Project 2

The second Project is about solving a classification problem for the cifar10 image dataset.
The project is also composed of 2 Tasks:

1. Create a Convolutional Neural Network classifier for the problem.
2. Hyper parameter tuning: Perform a grid search on the parameters below to find the optimal model, and then compare it with the model from Task 1

• learning rate: {0.0001, 0.01}
• number of neurons: {8, 64}

The code for each individual Task can be found in the src directory. To train the models, run the script through a terminal with the following command:

$ python3 file1.py

or

$ python3 file2.py

Each script creates a model and saves it in the deliverable directory. There, the command

$ python3 run_model.py

can be run to just see the comparison (using the T student statistic) of the two models, without having to train them.