The time of average, standard models is long gone. The world is getting more data-driven every day. This brings tremendous opportunities for the best predictive models to cut waste, delight customers, outdo the competition, and to make the world a better place. Based on this premise, Kaggle has seen rapid growth as a platform for competitive predictive modeling. Kaggle attracts the top minds to build algorithms for Allstate, Nasa, Wikipedia, Microsoft Kinect, Facebook, Merck, and others for large sums of money.
17:30 - 18:30 Drinks, Pizza
18:30 - 19:15 Kaggle war stories I - Tim Salimans
Tim Salimans is a Kaggle top contender who has won Kaggle competitions, generally places high, and is in the running to win $ 3,000,000. Tim will introduce us to the world of competitive predictive modeling: a world of intrigue, ensemble learning, collaborative modeling, and late night hacking. Tim often wields the sword of Bayesian learning, but will also pick other techniques if needed.
19:30 - 20:15 Kaggle war stories II - Laurens van der Maaten
Our second speaker, Laurens van der Maaten, is a postdoc at the Delft University of Technology, and recently won Kaggle's Merck Molecular Activity visualization Challenge.
Laurens will also share his Kaggle war stories and talk about t-Distributed Stochastic Neighbor Embedding(t-SNE). t-SNE is a new dimensionality reduction routine that is well suited for the visualization of high-dimensional datasets. This method, that he pioneered, was used in the winning submission. Laurens will also shed light on how t-SNE compares to other dimensional reduction algorithms like PCA and SVD.
20:15 - 21:00 Drinks
Call for speakers: talk to us about your interesting Data Science projects!