Archive for February, 2015

Michael Jordan, who co-authored the first paper on Latent Dirichlet Allocation, talks about the relationship between frequentist and Bayesian statistics.

The New Yorker’s bio of Yitang Zhang, who recently proved the Bounded Gaps Conjecture — the most important mathematical discovery so far this decade.

Tali Garsiel’s “How Browsers Work” is an epic exploration of the internals of modern web browsers, and the culmination of years of research (including reading the Webkit and Gecko codebases). While almost 4 years old now, it is still super-relevant for any front-end developer. The web browser is your platform — get to know it better!

Feeling a little older? Maybe the universe didn’t start 13.8 billion years ago, maybe it was hanging around forever as a quantum potential waiting to make a big bang.

Alvaro Videla schools us on the origin of harmful gotos and other programming myths.

Happy Friday!


Schrodinger’s unlucky Cat can be alive! The Cat has been used to represent possible quantum states of a particle. In Schrodinger’s conceptual model, the unlucky Cat is placed in a container with a vial of cyanuric acid that breaks open if the container is opened. It is thus assigned both dead and alive states since we are unable to observe it’s state without opening the container. Similarly, quantum particles are assigned multiple states, represented by probabilities, and cannot be determined because the act of measuring perturbs the state. Berkeley researchers have in part been able to tease out this uncertainty in part by accurately predicting quantum particle trajectories. Synopsis here.

Tired of manually choosing mates through Tinder? By analyzing a small set of previous decisions about which people he liked, Justin Long constructed an averaged face that met his tastes and automatically liked/disliked future pictures. With this tool, you can rest easy knowing that no more potential booty calls will slip through your fingers.

Happy cursed Friday everyone!

Don’t Use Bandits. Multi-armed bandit testing is very popular these days, but it’s critically important that it be applied correctly. Chris Stucchio does a great job of summing up some of the caveats that needs to be kept in mind when applying bandit algorithms.

Jeff Carter postulates that large banks can and should be replaced by smaller companies and give he gives several examples of startups that are doing finance much better than most of the big players.

The best computer vision algorithms can classify objects in real world imagery with impressively high accuracy. But what if you were to feed them carefully crafted garbage? Researchers used evolutionary algorithms to concoct a set of abstract images that tricked the best computers. (Read the summary at KDnuggets if you don’t have time for the paper.)

Happy Friday!