Machine learning introductions for beginners

[agitter]

This Nature Methods editorial about machine learning in biology could be a good resource for us to send participants before the workshop, though it's fairly heavy on deep learning. https://doi.org/10.1038/s41592-019-0432-9

[dltreu]

As a non-computation person, I thought it a good read, a broad look at the topic. It could also be a good introduction to the "ethics" questions that have come up in the pilots - responsible use of ML, biased data, etc.

[agitter]

Three pitfalls to avoid in machine learning is written at an appropriate level for our audience. Their pitfalls are all relevant for biology:

• Splitting data inappropriately
• Hidden variables (batch effects)
• Mistaking the objective

The article also links to Google's AI education site.

[cmilica]

this is great! Thank you!

[agitter]

This comic is a good overview of the major ML concepts we want to teach: https://cloud.google.com/products/ai/ml-comic-1/

[dltreu]

Excellent!

[agitter]

Here's another great article for beginners: How to Read Articles That Use Machine Learning: Users’ Guides to the Medical Literature

Some aspects toward the end are clinically oriented and less relevant for us, but most of the definitions and motivations are perfect.

[dltreu]

Tony, this is superb. A glossary, case studies and a good general introduction to machine learning.

It might be a piece to handout before the session (or between parts 1 and 2, if the new pilot is a two part), giving participants time to digest and develop questions.

Debora

[agitter]

This slide deck on evaluating machine learning claims is targeted toward a similar audience. The examples of which reported accuracies you would trust is a good illustration of how to evaluate models: https://www.slideshare.net/hoffmanlab/evaluating-machine-learning-claims-229405631

[agitter]

This repository is collecting interactive ML demos: https://github.com/MilesCranmer/awesome-ml-demos