parrisj

Just a couple of notes 1) The linked version is out of date here the most current version as of Nov 2011 http://web4.cs.ucl.ac.uk/staff/D.Barber/textbook/211111.pdf

2) @reader5000 It's legit he links to it from his homepage http://web4.cs.ucl.ac.uk/staff/D.Barber/pmwiki/pmwiki.php?n=...

This idea reminds me of ohlife. An email based journal service where every day they randomly choose a previous entry to send you. Its a cool idea. I just don't see how you'll get people to repeatedly visit.

Wow, I'm torn between laughter and sadness. Laughter because svm's are used everywhere, seriously google scholar returns 1.5 million papers mentioning support vector machines. Good luck suing everyone of those. And sadness, although unlikely, this could have a huge chilling effect on Machine Learning, Computer Vision, NLP, etc.

I'm not sure of all the details. I recently organized a face detection competition. And we included a couple of commercial face detectors for comparison. My boss informed me that since we purchased them under an education discount program we had to refer to them as "Commercial Competitors".

My view might be slightly skewed since the basis of the competition was to make it hard (low light, noise, occlusion, pose, atmospheric blur) and the commercial competitors are trained on easier datasets (close range, well lit, uniform pose and expression). While L1's system would perform better on driver's license images false positives are going to occur in any system. So when it flags a person it should be regarded as something fishy might be going on instead the oracle says this person is a terrorist who buys alcohol for teens, lock him away!

Part of this problem is their using software L1. L1 has terrible performance but their EULA states that you can't refer to them by name in Academic papers. So their always refereed to anonymously as the"Commercial Competitor".