Learning Machines

Taught by Patrick Hebron at ITP, Fall 2016

Week 7:

Looking Ahead:

In the second half of the semester, we will continue to investigate more elaborate machine learning algorithms while also starting to use a wider range of tools. These two elements as well as in-class workshopping and discussions about machine learning applications will help us to work towards final projects.

In class today, we will:

• Get a basic understanding of what Docker and LaunchBot do.
• Take a brief tour of the Jupyter notebook format.
• Discuss the programming model used in TensorFlow.
• Work through a few preliminary TensorFlow tutorials (including a word2vec example).
• Discuss embedding spaces in Self-Organizing Maps, t-SNE and word2vec.

Hello, TensorFlow:

This week, we will start to explore TensorFlow, using Docker and LaunchBot to streamline our workflow.

If you haven't done so already, please:

• Sign-up for a LaunchBot account.
• Follow the steps listed in the LaunchBot Getting Started document.

Once LaunchBot is running, in the "Clone an external project..." panel, enter the url for our class LaunchBot repo:

https://github.com/Hebali/launchbot-learning-machines.git

In LaunchBot's "Projects" tab, you should now see the launchbot-learning-machines project listed. Select this project and hit "Launch."

The first time you launch a project, it may take awhile to download and build the image.

Homework:

Assignment:

• Spend some time exploring the TensorFlow tutorials in our class LaunchBot and from the other sites linked there. Particularly, Aymeric Damien's TensorFlow-Examples, which includes many examples already formatted for Jupyter notebook.

Research Starting Points:

• CreativeAI
• arXiv Machine Learning
• arXiv Neural and Evolutionary Computing
• arXiv Artificial Intelligence
• Reddit Machine Learning
• Reddit Artificial Intelligence
• Deep Learning News

Readings:

• For this week's reading, spend some time exploring the sites linked above.