THREE years after online DVD rental firm Netflix offered a to anyone who could improve its movie recommendation system by 10 per cent, an international team of computer scientists has made the grade.
To take part in the competition, dozens of teams have been training their algorithms using records of the way some 480,000 customers rated 18,000 films. The algorithms are then tested on another data set of customer ratings which are known only to the Netflix judges.
Several teams have been closing in on the target. In December 2008, Netflix awarded a $50,000 鈥減rogress prize鈥� to a team called BellKor in BigChaos for achieving a 9.44 per cent improvement over the Cinematch system Netflix has been using till now.
Advertisement
That group recently joined forces with four others. On 26 June the combined team, which calls itself , announced it had achieved a 10.05 per cent improvement.
The prize money isn鈥檛 guaranteed yet. The rules of the competition give other competitors 30 days to beat the new figure before the winner can be officially named.
