Tera-Scale Machine Learning in Computational Advertising

Date/Time:
Thursday 12 February 2015, 6.00pm - 8.00pm

Venue:
The Davidson Building, 5 Southampton Street, London WC2E 7HA. The nearest underground stations are Covent Garden and Charing Cross.

Speaker:
Bruno Fernandez-Ruiz

Summary:

The growth of computational advertising over the past decade as a new scientific and industrial discipline has opened exciting statistical and machine learning challenges. We see training sets grow to billions of samples; we see the need for online learning of statistical models that accurately capture the dynamics of the real-time bidding and programmatic buying; we see the dimensionality of the models explode from hundreds to billions of parameters; we see the optimization problems dramatically increase in computational complexity and storage needs, with non-convex and non-linear problems becoming more widely spread. In this talk, we’ll walk through some of the computational optimization challenges we see in the state-of-the-art machine learning infrastructure, and we’ll introduce some of the research and applied directions we are taking to solving these problems.

Bio:

He is responsible for the development of Yahoo’s native advertising platform. Bruno’s teams use machine learning, big data, and search technology to infer knowledge out of data and blend ads and content in ways uniquely relevant to users. Prior to this role, Bruno was VP of Ad Targeting and Personalization at Yahoo, and prior to that, chief architect for Yahoo!’s cloud and platform group.

Before joining Yahoo, Bruno founded OneSoup, a mobile messaging startup, and he was a consultant in the Andersen Consulting Centre for Strategic Research in Sophia Antipolis, France.

Bruno received his M.S. degree in Transportation and Operations Research from MIT, and his B.S. and M.S. degree in Computational Mechanics for Structural engineering from the Universidad Politécnica de Madrid.

Admission:

Free. To gain admission please email your name to our Membership Secretary, Algirdas Pakstas, at algirdas.pakstas@gmail.com in advance of the meeting, including the title of the event and your name in the subject line of your email. Attendance lists will normally be finalised on the Monday preceding each meeting but late admission may be accepted by signing in to the Davison Building as a visitor.

Presentation

PDF Icon Tera-Scale Machine Learning - Bruno Fernandez-Ruiz