I saw a couple of stories in the news about machine learning this week so I thought I’d share them.
- Sifting through long, repetitive video to extract the highlights
- A new toy for bird-watchers: using facial recognition to identify birds
- Aggregating data from seven domains including brain scans, personality tests, demographic information, and genetics to predict teen binge drinking behavior
My favorite part of the teen drinking study is that they pulled together complicated data from a variety of sources, but in the end they used a fairly simple model (penalized logistic regression, with cross-validation to tune the model) that can be implemented in any standard statistical software package.