Commercial Applications of Machine Learning

Wednesday, November 4, 2015 - 17:30
TH 331
Krishnan Raman (Marin Software)

We delve into commercial applications of machine learning & computer vision at Marin, by examining 3 verticals - Search, Social & Display.
In Search, we use Decision Trees with pruning to improve Revenue Per Click prediction on search keywords with tail terms, across major publishers ie. Google, Bing, Yahoo.
In Display, we use Alternating Least Squares matrix factorization to compute user-user lookalike audiences for a retargeting platform.
In Social, we build a scalable library of TV ad-spots by computing robust image descriptors invariant to scaling, translation & luminance.


Krishnan Raman is a Principal Data Scientist at Marin Software. Previously, he has worked as a Data Scientist/Quant/Programmer at Twitter, Bank Of America, Sun Microsystems, Goldman Sachs and a bunch of startups. He has graduate degrees in CS, Math & Financial Math, and enjoys dabbling in derivatives on the side.