Carla Gomes app making national news

A team led by Carla Gomes and the Institute for Computational Sustainability that she directs has developed cell phone apps that herders in Kenya are using to track herds and report conditions at grazing sites; the information that is gathered complements satellite imagery. Computer Magazine and Futurity both picked up the article.

In a Cornell Chronicle article, PhD student Yexiang Xue comments on the optimization-problem aspects of trying to incentivize participants to report on under-visited locations, and Gomes notes, "If we can use sophisticated technology to run Amazon, we can use sophisticated technology to help people in Africa".

Date Posted: 2/26/2015
Permalink

Tool for helping tweets get more retweeted makes the news.

PhD student Chenhao Tan created an online tool that predicts which of two differently-worded versions of a tweet will get more retweeted.  The website, https://chenhaot.com/retweetedmore/ , has been discussed by CBS News, The Washington Post, Fast Company, and the Daily Mail.  The underlying technology is based on joint work with Lillian Lee and CS PhD alum Bo Pang.

Fast Company had this to say:  "It may not exactly be a writing guide on the level of the classic The Elements of Style, but as a guide for getting noticed online here in 2015, it’s certainly not bad."

 

Date Posted: 2/26/2015
Permalink

Kozen honored for theoretical computer science work

Dexter Kozen has been named one of the five 2015 Fellows of the European Association for Theoretical Computer Science (EATCS) for his scientific achievements in the field of Theoretical Computer Science. Kozen's citation is for "pioneering and seminal work in fields as diverse as  complexity theory, logics of programs, algebra, computer algebra and probabilistic semantics".

Cornell Chronicle highlights Kozen's acheivements here. 

Date Posted: 2/26/2015
Permalink

Smithsonian lists Prof. Saxena as one of the 8 innovators to watch in 2015, for his ongoing work on RoboBrain.

Ashutosh Saxena was named one of Smithsonian Magazine's "Eight Innovators to Watch in 2015" and was highlighted in the Cornell Chronical.  The article notes, "Ashutosh Saxena envisions a world where robots can heed commands, such as "pour me a coffee" or "load the dishwasher," without step-by-step instructions. But unlike the novelists and screenwriters who have also dreamt this, he is actually making it happen.

The Cornell roboticist and his team are building RoboBrain, a massive online search engine of sorts for robots to use to acquire the knowledge needed to understand and then perform a task. When posed a question, RoboBrain will crawl the Internet and word, image and knowledge databases for relevant information that the robot can digest."

Date Posted: 2/18/2015
Permalink

Keynote and outstanding reviewer awards at WSDM 2015

Thorsten Joachims gave a keynote address on "Learning from user interactions" at the Eighth ACM International Conference on Web Search and Data Mining (WSDM), one of the leading conferences on search and data mining on the Web.  He discussed how intelligent systems that learn from their users can be designed in a principled way. To view the Abstract click here.   To view Slides click here.

Also, graduating PhD student Karthik Raman was named a recipient of an Outstanding Reviewer Award at the same conference

Date Posted: 2/11/2015
Permalink

Xanda Schofield wins a Microsoft Research Graduate Women's Scholarship

Xanda Schofield has won a Microsoft Research Graduate Women's Scholarship. The award supports the second year of each winners' graduate study, and only ten are given annually, to first-year female graduate students worldwide in the areas of computer science, electrical engineering, mathematics, bioinformatics, and information science. (Schofield's win does not preclude eligibility for the Microsoft PhD Fellowship, which is open only to third- and fourth-year PhD students.)

Schofield's research interests are in topic modeling, with specific interests in phenomena in corpora with significant spelling or grammatical variation (such as Early Modern English or Twitter shorthand), and creating scalable and widely-accessible tools for topic modeling.

Date Posted: 2/05/2015
Permalink

New computation method helps identify functional DNA

 Nature Genetics published a paper by CS PhD student Brad Gulko, Cornellians Melissa J. Hubisz and Ilan Gronau, and CS field member Adam Siepel on a new computation method that helps identify functional DNA. The proposed model can be thought of as belonging to a class of nonparametric generative models, selecting a relatively low dimensional covariate space. In addition to being a competitive or superior predictor, says Gulko, "Our method allows other scientists not only to use the results, but to readily understand them.”

 The paper was highlighted on the Nature Genetics homepage.  

Date Posted: 2/03/2015
Permalink

Gün Sirer was quoted in a "Top Story" of Forbes Magazine

Gün Sirer was quoted in a "Top Story" of Forbes Magazine about the ongoing trial 
of Ross Ulbricht for running the Silk Road underground market under the Dread 
Pirate Roberts (DPR) alias. Ulbricht admitted to being an erstwhile operator of 
Silk Road, but pointed to the Mt. Gox CEO Mark Karpeles as the "real" DPR.
 

Date Posted: 1/21/2015
Permalink

How to Spoof Deep Learning Featured in Wired and MIT Technology Review

PhD student Jason Yosinski's joint work on how deep-learning algorithms can be easily tricked has attracted more press attention, scoring writeups in Wired and MIT Technology Review:

"A technique called deep learning has enabled Google and other companies to make breakthroughs  in getting computers to understand the content of photos. Now researchers at Cornell University and the University of Wyoming have shown how to make images that fool such software into seeing things that aren’t there."

"...the group generated random imagery using evolutionary algorithms.

Essentially, they bred highly-effective visual bait. A program would produce an image, and then mutate it slightly. Both the copy and the original were shown to an “off the shelf” neural network trained on ImageNet, a data set of 1.3 million images, which has become a go-to resource for training computer vision AI. If the copy was recognized as something — anything — in the algorithm’s repertoire with more certainty the original, the researchers would keep it, and repeat the process.... Eventually, this technique produced dozens images that were recognized by the neural network with over 99 percent confidence. To you, they won’t seem like much. A series of wavy blue and orange lines. A mandala of ovals. Those alternating stripes of yellow and black. But to the AI, they were obvious matches: Star fish. Remote control. School bus."

Date Posted: 1/14/2015
Permalink

Pages

Archived News:

more