UCLA Department of Statistics Seminar Series
Thu, 9/4/2014, 12:30 PM—1:30 PM
5264 Boelter Hall
Yosihiko Ogata and Jiancang Zhuang
Institute of Statistical Mathematics, Tokyo
Thoughts about space-time point processes and earthquake forecasting
In an interview format, the speakers will give their thoughts on the past, present and future of earthquake forecasting, and the use of point process methods and models. The event will be moderated by Rick Schoenberg of UCLA Statistics.
Paper Analyzing Effect of Images in the Media is Reported in the BloombergView
Congratulations to Jungseock Joo, Weixin Li, Francis F. Steen, and Song-Chun Zhu as their work, “Visual Persuasion: Inferring Communicative Intents of Images”, was both presented at IEEE CVPR 2014 and reported in the BloombergView. Their paper analyzed huge numbers of political images in the media in order to determine their likely effects on audiences. It brings together the tools of statistics and machine learning to analysis in political science.
Please see the article in the BloombergView at:
2014 UCLA DataFest is Featured in the Blog fivethirtyeight.com
See this link: http://fivethirtyeight.com/datalab/the-students-most-likely-to-take-our-jobs/
Judea Pearl has been elected to the National Academy of Sciences
Congratulations to Judea Pearl, Professor Emeritus who was elected in April to the National Academy of Sciences. This is one of the highest honors possible for an academic scholar in science, engineering or medicine. It rewards distinguished and continuing achievements in original research.
UCLA Statistics is #8 in QS World University Rankings
UCLA Statistics was ranked # 8 worldwide amongst Statistics & Operational Research programs in 2014. Details are available at:
The department was ranked #14 in 2013. Let's make this a trend!
More Innovation from VCLA
Some members of our VCLA (Vision, Cognition, Learning and Art) research group (from left to right: Yibiao Zhao, Yixin Zhu and Steven Holtzen) are building a cognitive robot, which will not only navigate and recognize objects, but also make sense of the world like a human and answer meaningful questions beyond “what is where”. Answers relating to:
- Functionality: “What is the object used for?”
- Physics: “How likely will the object fall if someone bumps the table?”
- Intentionality: “Why the man left the room without closing the door? Will he come back?”
- Causality: “What did the man do if he come back with a cup of coffee?”
A central goal of computer vision is to create computational systems whose visual recognition and scene understanding accuracy is comparable to, or better than, that of biological vision. Over almost 50 years, the recognition of explicit visual patterns, like face and handwritten digits, has matured to the point of being ubiquitous in modern industrial and consumer products. However, computers still cannot perform many important tasks that are trivial for human vision. Recognizing man-made objects that vary greatly in appearance and shape, but not in function, such as chairs, and anticipating physical dangers, are examples of such tasks. This gap between people and computers exists because computers cannot use rich common sense knowledge about the functional, physical and social mechanics of the world in the ways that people can.
The group aims to close this gap by exploring methods for representing and exploiting common sense world knowledge about function, physics, intentions and causality in ways that can make human-level performance possible in machines. By treating function and physics as principal determinants in how a visual scene is organized by man or by nature, and by describing the relationships between visual entities and ongoing events using a rich notion of causality, computer vision systems can go beyond labeling “what is where” in an image to building a sophisticated understanding of a scene’s three-dimensional organization over time, its physical dynamics, and what actions it affords. In effect, these abilities allow a robot to answer an almost limitless range of questions about an image using a finite and general-purpose model. This may ultimately be instrumental to designing computers that can pass the visual Turing test.
A workshop "Vision meets cognition: functionality, physics, intentionality and causality" is being co-organized by Yibiao Zhao in conjunction with CVPR 2014. This will bring together researchers from different sub-communities within computer vision, computer graphics, robotics and cognitive science, to illuminate and broaden interdisciplinary awareness and collaboration.
Two UCLA Statistics Faculty Win Prestigious "PAMI Helmholtz Test-of-Time Award"
Congratulations to Song-Chun Zhu and Alan Yuille who won the “PAMI Helmholtz Test-of-Time Award”. This was awarded by the IEEE PAMI Technique Committee at the International Conference on Computer Vision (ICCV) held in Sydney in December 2013. They received this honor for the paper entitled, “Region Competition: Unifying Snakes, Region Growing, and Bayes / MDL for Multiband Image Segmentation”, which they co-authored in 1996. The paper developed a new algorithm called “Region Competition” which first linked statistical models of images to partial differential equations (PDEs) for image segmentation. The paper that wins the “PAMI Helmholtz Test-of-Time Award” is one that is frequently cited by other papers in computer vision, judged to have made a powerful impact to the field and must have been published by ICCV more than a decade ago.