A number of statistical and machine learning techniques have "gone mainstream" in astronomical research right now, so it has made a very exciting time. There is

A number of statistical and machine learning techniques have "gone mainstream" in astronomical research right now, so it has made a very exciting time. There is even a good textbook out now on the subject, Statistics, Data Mining, and Machine Learning in Astronomy: A Practical Python Guide for the Analysis of Survey Data by Ivezic that we used for a graduate seminar series. This mostly focuses on Bayesian inference and simple techniques like clustering algorithms, support vector machines, etc. and not super fancy "deep convolutional neural networks" like I see on HN most days :) A funny anecdote makes me wonder actually if there have been barriers to uptake in astronomy. You would think huge datasets and low manpower would be the perfect recipe for machine learning! Well, historically cheap labor could solve this by throwing undergrads at this. I actually even categorized ~10,000 maps of interstellar clouds by eye for one project (it took about a week if you don't count the time to write the GUI). The anecdote though was that one of the other grad students in the department was telling me how he was working on applying machine learning techniques to classifying galaxy morphological types. They had somewhere around 100,000 galaxies to type from their survey. But, they first applied to the citizen science organization Galaxy Zoo, and were featured as a project. They submitted classification instructions and the data, and it was completed in... less than 24 hours. He told me afterwards that, with how easy and fast it was, why bother spending weeks writing and tuning some neural network yourself? This made my wonder whether if Galaxy Zoo and other citizen science projects are such good resources, would it stymie the need for incorporating machine learning techniques into most projects? There are huge projects like the SKA (mentioned in the article) and LSST that will give torrents of data. Those might be too large for citizen science, and there are advantages to objective statistical methods to begin with.