Python!

I’m a big fan of Python for Machine Learning, and Jupyter Notebooks as teaching tools.  You can (and should) use the tools that work for you, but the following files are based on Jupyter Notebooks (.ipynb files).  Some of these files are included, or you can copy/paste code into your notebook.

• We will be using Jupyter Notebooks (.ipynb), running Python3
• You will need an Anaconda distribution.
• The packages we use are included in Anaconda.
• You may need to update certain packages
• You can install and update packages via the pip installer or through the conda installer.
• Google is your best friend when it comes to making things work on your own computer.
• I’d recommend updating sklearn (sci-kit learn), the main machine learning library to at least version 1.0.  This will give you the ability to create visualizations of your algorithms, which is not included in version 0.**
• pip install –upgrade scikit-learn, or depending on your permissions:  pip install –upgrade scikit-learn -user

I find the pip installer more straightforward than conda – but it doesn’t matter as long as it works for you.  Use these commands in the AnacondaPrompt.

Alternatives to running anaconda on your laptop:

• BCIT students: Workspace / Apps Anywhere lets you access a BCIT computer.  You may need a VPN to use from home.
• BCIT students: The oddly named  syzygy server at https://bcit.syzygy.ca/ lets you login with your BCIT credentials.  This lets you run programs off of the BCNet servers (physically located at UBC, but available to post-secondary institutions over BC)
• You can upload a jupyter notebook (.ipynb) to a Google Drive, and run it through Google Colab.  This is free at the base level, and lets you run off of Google servers.  No privacy guarantees here!  And you may need extra code in your first block to install needed packages.
• Other cloud servers such as AWS may or may not be free.  AWS gives students a limited amount of free credits – I’d save those for training a big neural network
• Hate Anaconda and want to use your own python distribution?  That’s fine, but you may lose the ability to run one block of code at a time.