T-809-DATA exercises

This repository contains exercises and supporting code for the course T-809-DATA at Reykjavik University. The assignments in the course are computer assignments and require at least some foundational knowledge in Python. All the assignments are written for Python and together require only one virtual environment for requirements. See the installation section for more information on that and best practices.

Modules

The repository contains the following modules:

• Assignment 0: Introduction
• Assignment 1: Decision Trees
• Assignment 2: K-Nearest Neighbors
• Assignment 3: Sequential Estimation
• Assignment 4: Classification Based on Probability
• Assignment 5: The Back-propagation Algorithm
• Assignment 6: Applying Neural Networks
• Assignment 7: Linear Models For Regression
• Assignment 8: Support Vector Machines
• Assignment 9: Random Forests
• Assignment 10: Boosting
• Assignment 11: K-Means and Gaussian Mixtures
• Assignment 12: Principal Component Analysis

We will add modules gradually throughout the course.

Installation

Installing Requirements

All the code in this repository is written in Python and requires Python 3 (any version after 3.6.9 should be ok). The recommended approach is to create a python virtual environment:

• Create the virtual environment with one of the following:
  • macOS/Linux: python3 -m venv env or virtualenv -p python3 env
  • Windows: python -m venv ./env or py -3 -m venv .env
  • Note: Make sure that python3 or python points to Python3.7+ interpreter. You can run python3 --version to see which version you are running. All python installations should be listed by running e.g. ls usr/bin/python. You can also use VS Code to discover all python interpreters on your OS, see the next section.
• Activate it with source env/bin/activate if you are in the project directory. Otherwise you do source /path/to/your/environment/bin/activate.

You can however use Python in any way you see fit and perhaps you may have all the requirements already installed system wide.

Install Python requirements with pip install -r requirements.txt. You can of course install any additional python requirements using pip, just make sure you have your virtual environment activated when you do.

Using VS Code + Python (Optional)

To get the best experience make sure that your VS Code workspace is using the correct Python interpreter. If you are using a virtual environment then the workspace setting python.pythonPath has to be set to /path/to/venv/bin/python. Normally VS Code takes care of doing this for you by recognizing that there is a virtual environment in the workspace. If not:

• Make sure you have the VSC Python extension installed (search for ms-python.python in the extension search)
• Press the settings cog in the bottom left inside VSC and select settings.
• Select Workspace
• search for pythonpath and edit the value to point to your python interpreter as explained above.

You can additionally see in the bottom toolbar of VS Code which python interpreter is running (it should say e.g. Python 3.6.9 64-bit ('env': virtualenv). By clicking that (or entering ctrl+shift+p and enter python interpreter in the prompt) you can select the python interpreter for the work space.

You can read a more detailed document about python environments in VSC here.

Markdown Math

You will notice that some of the READMEs contain something like:

$$p(x) = \frac{1}{\sqrt{ 2 \pi \sigma^2 }}e^{ - \frac{ (x - \mu)^2 } {2 \sigma^2} }$$

This is LaTeX style math markdown. Depending on the markdown viewer you use you might not see this rendered as math. To render this as LaTeX math inline in the markdown in VS Code, install the extension: Markdown Math. Then open the markdown file and click the icon with the magnifying glass over a book that pops up next to the filename in the editor. The rendered markdown appears to the right of your file.

Using this repository

Assignments are separated into folders. Each assignment folder contains the following files:

• README.md: The assignment description. For the best experience, open the file in VSC and press either ctrl+K V or the small magnifier glass icon on the right side in the header to Open preview to the side.
• template.py: Contains a template for you to fill in with your own code. This is the file to submit.

Each directory might also contain:

• example.py: Example code that is relevant to the assignment. Examples from this file are often referenced from the README.md
• tools.py : Sometimes we supply some helper functions for you to use. They will be found in this file.
• Directories that contain image, text or any other type of data relevant to the assignment.

How to turn in the assignments

All your code should be turned in as a python script, i.e.

• A single file with a .py ending. Although you can use a notebook in your development and then port your code over to template.py if that suits you.
• The file should contain all the functions with exactly the same function names as listed in the assignment description below
• Use the supplied template.py file to fill in your code
• Plots should be turned in as well. The naming convention for submitted plots is as follows: The Z-th plot under Section X.Y should be turned in as X_Y_Z.png

Questions based on independent work

Some assignments will have independent sections. In those assignments, the assignment itself counts 80% towards the assignment grade and the independent section 20%. The idea behind the independent questions is to allow you to use your intuition to add relevant insight to your assignment without clear instructions. Examples of this could be for example:

• Test your assignment code on a different dataset.
• Compare performance across different parameter configurations.
• Expand your model according to some hypothesis and compare performance to base model
• Test your model on different amount of training data and comment on difference in performance
• etc.

Assignment 1 and Assignment 2 have pre-formulated independent sections. Here the independent questions have already been formulated to give you an idea about what is suitable for the independent sections in future assignments.