Link to the paper mentioned in the article: http://people.csail.mit.edu/rinard/paper/popl16.pdf http://people.csail.mit.edu/rinard/paper/popl16.pdf quick summa

Link to the paper mentioned in the article: http://people.csail.mit.edu/rinard/paper/popl16.pdf http://people.csail.mit.edu/rinard/paper/popl16.pdf quick summary: They trained on apr, curl, httpd, libtiff, php , python, svn, wireshark. It basically trains the machine learning algorithm with the Abstrat syntax tree (AST) of the code before and after the successful patches. It then uses the machine learning predictions on real world defects and see what comes out of it.