— In the News —
Great longread about the origins of crowdsourced data at DARPA. It begins in Afghanistan’s only tiki bar, where patrons were given free beer in exchange for miscellaneous data that they uploaded onto a 1 TB drive, kept at the bar...
Cool research! This article from OpenAI explores how to train cooperative agents to develop their own language.
While AI systems can match many human capabilities, they take 10 times longer to learn. Now, by copying the way the brain works, Google DeepMind has built a machine that is closing the gap.
— Tools and Techniques —
Bare bones example of TensorFlow, the super popular machine learning package from Google. This is the simplest introduction you'll find.
Detailed article that presents several approaches for interpreting machine learning models and results. This goes well beyond the usual error measures and assessment plots.
Flickr recently introduced a new feature that helps users find photos that are similar to a selected image. Considering a variety of user expectations and the large number of photos involved, this is not a trivial task. This post describes the challenges and how Flickr approached them.
Nice overview of techniques and resources for strategizing your March Madness bracket predictions.
— Data Viz —
Facebook Research just released visdom, a new tool to facilitate visualization of remote data. It supports Torch and Numpy, which makes it well suited for scientific experimentation. It looks super useful and judging by the interest around the web this week, it's going to be popular too.
Solid intro to working with data in Pandas and then exploring it using Seaborn, a Python data visualization library.