Matt Gattis specializes in machine learning pertaining to recommendation engines. eBay acquired the recommendation engine startup he co-founded, hunch.com, where he was the CTO and responsible for R&D. Currently, he is continuing to develop recommendation engine technology for merchandising at eBay. Prior to hunch, he worked on heuristics for detecting web security threats at SiteAdvisor, which was acquired by McAfee. Prior to that, he graduated from MIT with a degree in computer science.
And a brief description of what he plans to discuss:
The discussion will cover the basics of using prediction modeling to generate recommendations. We’ll start with the simplest algorithms possible and move into the state of the art. This includes nearest neighbor classification, least squares regression, dimensionality reduction, and generative latent factor models. We’ll also go over some of the common problems that arise with data collected on the web, such as it’s size and inherent bias, and talk about ways to model around these issues and implementations that scale.