Spam is not only a universal irritant, but also a major drain on the resources of companies whose businesses are web-based. In today's age of account phishing, Nigerian princes, and botnets, tackling spam needs much more than a simple rule or keyword based classification system. In fact, thinking about spam classification often raises questions about some of the fundamental assumptions in machine learning and poses significant algorithmic scaling challenges. In this talk, we will discuss some of the novel algorithmic and machine learning questions motivated by spam.

About the speaker:
Anirban Dasgupta is currently a Senior Scientist at Yahoo! Labs, where he works on algorithmic problems for massive data sets. He did his undergraduate studies at IIT Kharagpur and doctoral thesis at Cornell University on learning mixtures of distributions using spectral techniques. At Yahoo!, Anirban has since worked on abuse prevention, large scale machine learning, scheduling, analysis of large social networks and randomized algorithms in general.