Ben Taskar passed away recently. It's a sudden and tragic loss for the machine learning community, and our deepest condolences go out to his friends and family. I've only known Ben professionally; we've chatted a few times over the years. By all accounts, he was a wonderful and kind person. More information can be found here.
I knew Ben mostly through his research, which has been exceptional throughout his career as well as an inspiration for my own work. The term "rising star" is reserved for a select few in the research community, and Ben most definitely deserved that moniker.
In graduate school, I literally began my machine learning career reading Ben Taskar's papers. Ben was a pioneer in the area of Structured Prediction. His thesis work on Max Margin Markov Networks was a revelation, and has stood the test of time as one of those foundational papers that people refer to over and over again. He's also done some great follow-up work extending it as well.
But Ben was only just getting started. More recently, his group has done some extremely elegant work extending the limits of what structured prediction models can be applied to. For example, his work on Structured Prediction Cascades (with David Weiss) was one of the first principled approaches for discriminatively learning efficient approximations of complex structured prediction models with learning-theoretic guarantees. As another example, his work on Structured Determinantal Point Processes (with Alex Kulesza) is quite possibly the most elegant way of building probabilistic models of redundancy that I've encountered thus far.
Every year for the past several years, I would regularly browse his website in anticipation of finding interesting new papers that his group has recently published. Although the shock I'm feeling must pale dramatically in comparison to that felt by his family and friends, it is nonetheless profoundly saddening that this great star in the machine learning community has seen his chapter end so abruptly and prematurely.