I haven't been following the prize competition all that closely, but I'm intending to read through some of the technical descriptions and see what I can learn.
In the end, there were apparently two successful approaches: BellKor's Pragmatic Chaos, and The Ensemble.
The Ensemble apparently got a better score on the test, but BellKor's Pragmatic Chaos got a winning score earlier.
Both groups have made detailed information about their algorithms and techniques available, which is wonderful!
It's interesting that both these teams were collaborative efforts of multiple smaller teams. This seems to be increasingly common in modern Big Science. As one of the researchers noted:
"Some people on our team think this is the way that problems will be solved in the future," said Chris Hefele (updated) of The Ensemble, one of the two qualifying teams. "Large problems, with large teams all across the globe collaborating by internet."
The field of so-called "artifical intelligence" has been around for a long time, and comes and goes in favor. Some of the modern techniques, such as neural networks, Bayesian statistics, etc. are now extremely well developed, and obviously are quite successful. Perhaps this success will lead to both renewed interest and renewed respectability for these fields.
Meanwhile, Netflix is so pleased with the results that they're already planning Prize 2!
No comments:
Post a Comment