Aug 25, 2015
Qualcomm products mentioned within this post are offered by Qualcomm Technologies, Inc. and/or its subsidiaries.
On September 15, Qualcomm Research, a division of Qualcomm Technologies, Inc., and the University of Amsterdam (UvA) will open a new joint research lab focused on advancing state-of-the-art machine learning techniques for mobile computer vision. Professors Arnold Smeulders and Max Welling, along with Assoc. Prof. Cees Snoek, will direct the new lab and approximately 15-20 PhD students and postdoctoral researchers.
Bringing machine learning together with computer vision—with an emphasis on mobile and embedded applications—will foster new approaches to more intelligence in smartphone cameras, robotics, automotive and Internet of Everything (IoE) applications. “This is the exciting era to bring vision—the most important of our senses—to mobile, unlocking a million new ways of looking at the world. To the new Qualcomm/UvA Lab, it is not just the excitement of understanding visual cognition but equally inspiring is the potential to make it work for billions," said Prof. Smeulders.
A renowned researcher in the field of deep learning—particularly in next-generation approaches such as semi-supervised learning and generative models—Prof. Welling added, "The key to next-generation deep architectures will be to learn representations that are not only good at discrimination or classification but also at [image] synthesis/generation, and to understand the transformational properties of those representations. This type of regularization will enable learning deep networks from much smaller datasets. Another potentially high impact direction in deep learning research is the combination of ideas from Bayesian learning, distributed learning and privacy preserving learning."
The potential for effectively learning from smaller datasets, while preserving user privacy and harnessing the potential of a distributed network of devices will be a game changer in the development of more intelligent devices.
In addition to advancing machine learning techniques for mobile computer vision, UvA’s proximity to the recently established Qualcomm Research office in Amsterdam—where Asst. Prof. Snoek is also a principal engineer—is expected accelerate the development of more academic-industrial collaboration opportunities.
Besides the opening of the research lab, Qualcomm Research is excited to announce that Yoshua Bengio, a professor of Computer Science at the University of Montreal and a world-renowned researcher in machine learning, is now a scientific advisor to Qualcomm Zeroth initiative. The Zeroth team will benefit greatly from the wealth of expertise and insight that Prof. Bengio brings—he’s one of the most frequently cited authors in deep learning research.
“The short-term opportunities for deep learning is the development of specialized hardware. This could be useful for two reasons: to train much more powerful models (which could take advantage of larger quantities of data), and to enable their use by more people on low-power devices,” said Prof. Bengio. “Regarding the long-term challenges, one that I consider most exciting is unsupervised learning, which would allow our models to expand the scope of their intelligence by allowing the computer, like a child, to learn not just from highly directed human guidance but also by observation and experience from the much larger set of data available online, which have not been curated and labeled by humans. In this quest, I am personally very interested in taking more cues from neuroscience, since there currently is a large gap between machine learning and our theories of how brains learn.”
Fittingly Prof. Bengio will be the guest keynote speaker at the inauguration ceremony for the Qualcomm-University of Amsterdam lab on Sept 15.
The combined efforts of the Zeroth Research team with guidance from Professor Bengio, and collaborations in place with Professors Smeulders and Welling along with Asst. Professor Cees Snoek, pave the path for what will be possible in mobile and embedded systems with vision-based understanding. Stay tuned for updates and exciting results from both Qualcomm Research and the joint lab as they happen!
Researchers interested in applying for positions in the joint lab should start here.
Read more about the lab and upcoming research projects here.