Sunday, January 12, 2014

Another Paris ML Meetup#7 talk: What does it take to win the Kaggle/Yandex competition

I mentioned on Friday the three talks we will be having on Wednesday for the Paris Machine Learning Meetup #7: Machine Vision: VMX, Atheer One and VLAD

We will also feature Kenji Lefèvre-Hasegawa from the 'Dataiku Science Studio' team  (that includes Paul MasurelMatt ScordiaKenji Lefèvre and Christophe Bourguignat ) which just won First Prize of the Kaggle/Yandex competition out of 194 contenders ( the team listed as number one is OOC, i.e. Out Of Competition as it is a Yandex team). At the previous meetup, a member of that team, Matthieu Scordia, presented us what the team's strategy was at the time. I am told that the strategy changed in the last month of the challenge so we will get more insight directly from the horse's mouth.

Title: What does it take to win the Kaggle/Yandex competition

Abstract: I'll give a feedback on the "personnalized web search" Kaggle challenge for which our team won First prize. I will focus the talk on both the experience itself (team work and tools) and the models we've used. 

No comments: