One of my favorite papers at this year's conference was Generalization Bounds and Consistency for Latent Structural Probit and Ramp Loss by University of Chicago’s David McAllester and Joseph Keshet, which analyzed the statistical consistency of losses for structured prediction problems (like speech recognition, machine translation, and visual scene parsing). This was one of those nice theoretical papers which proved good properties about effective algorithms that until now were only understood as heuristics.
After the conference, we headed to Spain's Sierra Nevada for the NIPS workshops. The workshops at NIPS provide researchers a chance to focus on more specific communities, and Googlers helped to organize five workshops this year on various topics, from learning semantics to big learning. Several Googlers had accepted papers and talks at other workshops as well, such as Samy Bengio and Gal Chechik, who spoke at the workshop on Beyond Mahalanobis: Supervised Large-Scale Learning of Similarity.
I spent most of the second day at my own workshop on Domain Adaptation, which I co-organized with Googlers Corinna Cortes and Afshin Rostami. Our workshop brought together theoretical and practical domain adaptation, and the invited speakers included Shai Ben-David and Mehryar Mohri from the theory side and Dan Roth from the applications side. Since my workshop was next door to Googlers Doug Eck and Ryan Rifkin's workshop on Machine Learning and Music, I got to hear some interesting musical demonstrations from that workshop, as well. In addition to the Googler-run workshops, I really enjoyed the workshop on Language and Vision, which featured invited talks by Google postdoctoral fellow Percy Liang on the pragmatics of visual scene description and Josh Tenenbaum on physical models as a cognitive plausible mechanism for bridging language and vision.
As the workshop weekend drew to a close, some NIPS attendees scrambled to get home for the holidays in light of an airline strike. We hope the skies look clear for next year when NIPS lands in Google’s neck of the woods, Lake Tahoe!
John Blitzer, Research Scientist
Google at NIPS 2011
Googlers co-authored the following papers:
- Algorithms and hardness results for parallel large margin learning by Rocco Servedio, Philip Long
- Co-Training for Domain Adaptation by Minmin Chen, Kilian Weinberger, John Blitzer
- Learning large-margin halfspaces with more malicious noise by Philip Long, Rocco Servedio
- Big Learning: Algorithms, Systems, and Tools for Learning at Scale by Joseph Gonzalez, Sameer Singh, Graham Taylor, James Bergstra, Alice Zheng, Misha Bilenko, Yucheng Low, Yoshua Bengio, Michael Franklin, Carlos Guestrin, Andrew McCallum, Alexander Smola, Michael Jordan, Sugato Basu
- Domain Adaptation Workshop: Theory and Application by John Blitzer, Corinna Cortes, Afshin Rostamizadeh
- Learning Semantics by Antoine Bordes, Jason Weston, Ronan Collobert, Leon Bottou
- Sparse Representation and Low-rank Approximation by Ameet Talwalkar, Lester Mackey, Mehryar Mohri, Michael Mahoney, Francis Bach, Mike Davies, Remi Gribonval, Guillaume Obozinski
- International Workshop on Music and Machine Learning: Learning from Musical Structure by Rafael Ramirez, Darrell Conklin, Douglas Eck, Ryan Rifkin
- Online Similarity Learning: From Images to Texts by Samy Bengio
- Improving the speed of neural networks on CPUs by Vincent Vanhoucke, Andrew Senior, Mark Z. Mao
- Reading Digits in Natural Images with Unsupervised Feature Learning by Yuval Netzer, Tao Wang, Adam Coates, Alessandro Bissacco, Bo Wu, Andrew Y. Ng
- Results from a Semi-Supervised Feature Learning Competition by D. Sculley
Note: Googlers in blue.
Add a comment