Laziness is the mother of invention. Some people, mainly the CVPR reviewers, were trying to use AdaBoost to classify good papers from bad ones just by judging their layouts and formatting. The algorithm from the paper (Von Bearnensquash, 2010) can achieve a true negative rate of over 50% with a false negative rate of about 15%. The classification of a paper needs only 0.5s.
It was an informative read for me, especially for the interesting observations about what makes a good (or bad) paper. It’s funny that to be classified as a qualified paper for CVPR, the authors intentionally added a set of Maxwell’s equations to it. Weird but effective!
Check for a detailed review of another great example of AdaBoost applications.
References
Von Bearnensquash, C. (2010). Paper gestalt. Secret Proceedings of Computer Vision and Pattern Recognition (CVPR).