What should a theory of Big Data say?
I checked out my new apartment in Berkeley a few days ago. My very senior landlord asked me to explain to him what it is that I do. So, I said: “I’m a computer scientist.” He gave me an empty stare as...
View ArticleThe Zen of Gradient Descent
Ben Recht spoke about optimization a few days ago at the Simons Institute. His talk was a highly entertaining tour de force through about a semester of convex optimization. You should go watch it. It’s...
View ArticleHello TCS Aggregator!
I’ve finally managed to get on the TCS aggregator! That is under the condition that I use a special tag indicating that my post may be mathy enough for the greater aggregator community. Since I’ve been...
View ArticleThe Geometric View on Sparse Recovery
Sparsity is of fundamental importance in much of signal processing, optimization, computer science and statistics. It was also a major theme in both workshops so far in the Simons Big Data program. The...
View ArticlePower Method still Powerful
The Power Method is the oldest practical method for finding the eigenvector of a matrix corresponding to the eigenvalue of largest modulus. Most of us will live to celebrate the Power Method’s 100th...
View ArticleFalse Discovery and Differential Privacy
The Simons program on Big Data wrapped up about a month ago with a workshop on Big Data and Differential Privacy. With the possible exception of one talk on the last day after lunch, it was a really...
View ArticlePearson’s polynomial
A 120 year old algorithm for learning mixtures of gaussians turns out to be optimal So, how was math writing in 1894? I’d imagined it to be a lot like one of Nietzsche’s last diary entries except with...
View ArticleRobustness versus Acceleration
My blog post on the Zen of Gradient Descent hit the front page of Hacker News the other day. I don’t know how that happened. It got me more views in one day than this most humble blog usually gets in...
View Article
More Pages to Explore .....