Introduction to Modern Machine Learning with Neural Networks

Dr. Mohammad M. Khajah
Associate Research Scientist
Systems and Software Development Department (SSDD)

This repository contains all the materials for my KISR course on neural networks.

1. Installing Python

We will use Miniconda, a free minimal installer of Python and its related packages.

Windows

Download and run the Miniconda 3 Windows 64-bit installer. Don't change the default options in the installer.

Mac OS

For Intel Based Macs Follow the instructions to install Miniconda on Mac OS.

For M1 Based Macs Download the community-supported version of miniconda, known as miniforge, then install from the Terminal application as follows:

bash Miniforge3-MacOSX-arm64.sh

2. Configuring the Environment

1. In Windows, launch the Anaconda Prompt (miniconda3) application (open the start menu and just start typing the application name until you find it), or Terminal in Mac OS.

2. Create a conda environment named nncourse.

conda create -n nncourse

A virtual environment lets developers isolate different versions of the same package. Python packages add extra functionality, such as plotting diagrams, machine learning, linear algebra, web development, etc.

3. Activate the nncourse environment.

conda activate nncourse

4. Install the required packages for this course into the nncourse environment.

For non-Apple M1 machines:

conda install numpy pandas matplotlib scikit-learn tensorflow

For Apple M1:

conda install -c apple tensorflow-deps
python -m pip install tensorflow-macos
conda install pandas matplotlib scikit-learn

numpy is the most fundamental library for math, vector, and matrix operations. pandas makes it easy to interact with tabular data (csv and excel files). matplotlib is used for plotting diagrams. tensorflow lets us build neural network models.

5. Install the jupyterlab package from the third-party package repository known as condaforge:

conda install -c conda-forge jupyterlab

jupyterlab is an interactive web-browser based layer on top of Python that makes it easy to mix rich documentation and code.

6. You can close the terminal now.

3. Running Jupyter Notebook

All coding in this course will be done within the jupyter notebook. To launch it:

1. Launch the Anaconda Prompt (miniconda3) application in Windows or Terminal in Mac OS.

2. Activate the nncourse environment you created earlier.

conda activate nncourse

3. Launch jupyter lab:

jupyter lab

This will launch a local web server running Jupyter and then it will run a web browser and point it to the local server's address. If the browser does not run, copy the URL that appears in the console and paste it into a web browser.

Do NOT close the prompt/terminal application while jupyter is running, as you'd kill the local Jupyter server.

4. You can now navigate your local file system via the left side pane in the browser.

4. Downloading and Running Course Notebooks

1. On the course's Github page, click the green "Code" button.

2. From the drop down menu, click "Download ZIP".

3. Uncompress the downloaded file into a convenient location (e.g., the desktop).

4. Within the open jupyter lab tab in your web browser, navigate to the location where you uncompressed the file.

5. Double click on one of the .ipynb files in the left pane to open it in the right pane.

6. Jupyter notebooks consist of cells. Each cell can be Python code or Markdown (documentation) or plain text. Click on the play button at the top of the notebook to execute the current cell and move to the next one. You can also press Shift+Enter to do this.

Resources

• Python Numpy Tutorial: http://cs231n.github.io/python-numpy-tutorial/

• Stanford's CS231N class: http://cs231n.github.io/