Cal Berkeley EECS professor David Patterson did a great job this year organizing Berkeley’s EECS Annual Research Symposium on Thursday, February 23rd.
B.E.A.R.S. used to always be held in a theater in an engineering building on campus, however last year it was moved to International House to accomodate more people. But it was crowded and dark in there, and not a good fit.
This year, B.E.A.R.S. was held in Pauley Ballroom instead, atop the Student Union next to Sproul Plaza, and between great weather and plentiful refreshments, the audience of about 400 people seemed happier than ever to be there.
Good vibes in 2012 could also be attributed to Patterson’s choice of presenters, and the time allotted to each. Instead of the usual 10+ professors, each with 20 or 30 minutes to describe their research (academics who’ve consistently felt entitled to run long), Patterson limited faculty speakers this year to just 3. Each of them spoke succinctly and, miracle of miracles, stayed within their allotted time.
Hence, the fourth and final part of the 2012 program started on schedule, and consisted of 11 grad students talking for 5 minutes each about their respective research and labs. It was great – totally enjoyable and informative.
The uber-topic at B.E.A.R.S. this year was Big Data at Berkeley, and the presentation with the Biggest Gravitas was pretty clearly Prof. Michael Jordan’s.
With joint appointments in EECS and Statistics at Cal, Jordan has combined his expertise in data analysis with his knowledge of Big Data to develop, in conjunction with others in his lab, a Bag of Little Bootstraps [BLB] for Massive Data, “a new procedure which incorporates features of both the [classic] bootstrap and sub-sampling to obtain a robust, computationally efficient means of assessing the quality of estimators.”
In other words, Jordan et al want to find a better way to slice and dice the cyber data we’re all drowning in. Important, because every conclusion we come to today is somehow based on analyzing that data. Jordan says a lot of those conclusions are wrong, however, because our analysis tools are misguided, so his BLB hopes to fix that.
Also speaking at B.E.A.R.S., Prof. David Wagner scared the bejezzus out of everybody with horror stories about maleficent malware and evil apps – but he also offered clever schema for analyzing mega-cyber-data to produce solutions for the likes of Google and RSA. Wagner said it’s in society’s best interest to help these guys solve their mega-security nightmares, because we will all benefit.
The third faculty speaker was Prof. Clair Tomlin, who talked about how Big Data in combo with Big Feedback Loops can be used to fly Big Quad-copters. Well, maybe they aren’t so big, but the future promise of these nifty/nasty little devices is pretty obvious – everything from security to search & rescue.
To illustrate the work coming out her lab, Tomlin showed video of a quad-copter catching ping-pong balls. You think that’s trivial? When’s the last time you wrote a set of dynamic equations to describe the motion of a quad-copter, the motion of a ping-pong ball, and a control system to link them up? The video can be seen below.
Before you watch it, however, note there are two additional videos here as well. The one from the University of Pennsylvania is pretty creepy, and includes a swarming cloud of quad-copters.
The other one, however, is a work of art and not out of anybody’s EECS lab. It’s the total creation of Bay Area artist/inventor Don Severns, who not only built his own tri-copter, but cobbled on the GoPro camera, did the flying, the shooting, the editing, and made the surf board seen therein. Enjoy!
Ping-pong balls at U.C. Berkeley …
Creepy swarms at University of Pennsylvania …
Masterful homage to the power of the place from Don Severns …