Kihyuk Sohn

Ph.D. in Electrical Engineering: Systems
University of Michigan, Ann Arbor

Research
I am a researcher in Media Analytics group of NEC Laboratories America since July 2015. Before joining NEC lab, I completed my Ph.D. at University of Michigan under the supervision by professor Honglak Lee. I have broad interest in machine learning and computer vision. Specifically, my research focuses on supervised and unsupervised deep representation learning with applications to computer vision, audio recognition, and text processing, using graphical models that are invariant to many factors of variation for robust perception from complex and multimodal data.


Education
Curriculum Vitae [pdf]

September, 2008 ~ June, 2015
    Ph.D. in Electrical Engineering: Systems, University of Michigan, Ann Arbor, MI
    Thesis title: Improved deep representation learning with complex and multimodal data.
    Thesis advisor : Professor Honglak Lee
March, 2003 ~ February, 2008
    Korea Advanced Institute of Science and Technology (KAIST), Daejeon, Korea
    B.S. in Electrical Engineering and Computer Science and Mathematical Science


Contact information
Email:
    ksohn [at] nec-labs [dot] com
    kihyuk.sohn [at] gmail [dot] com
Address:
    10080 North Wolfe Road, Suite SW3-350
    Cupertino, CA 95014


News
  • Our paper "Learning Structured Output Representation using Deep Conditional Generative Models" got accepted to NIPS 2015. [pdf][supplementary material][bib]
  • Joined Media Analytics group of NEC Laboratories America as a researcher on July 1, 2015.
  • Passed Ph.D. dissertation defense on June 11, 2015.
  • Our paper "Improving Object Detection with Deep Convolutional Networks via Bayesian Optimization and Structured Prediction" got accepted to CVPR 2015 as an oral presentation. [pdf][supplementary material][tech report][code] (OpenCV People’s Vote Winning Paper) [link]
  • Our paper "Improved Multimodal Deep Learning with Variation of Information" got accepted to NIPS 2014. [pdf][pdf (full)][bib]
  • Our paper "Learning to Disentangle Factors of Variation with Manifold Interaction" got accepted to ICML 2014. [pdf][bib][code]

Publications


[13] Understanding and Improving Convolutional Neural Networks via Concatenated Rectified Linear Units.
Wenling Shang, Kihyuk Sohn, Diogo Almeida, Honglak Lee
ArXiv preprint [pdf]

[12] Attribute2Image: Conditional Image Generation from Visual Attributes.
Xinchen Yan, Jimei Yang, Kihyuk Sohn, Honglak Lee
ArXiv preprint [pdf]

[11] Discriminative Training of Structured Dictionaries via Block Orthogonal Matching Pursuit.
Wenling Shang, Kihyuk Sohn, Honglak Lee, Anna Gilbert
In SIAM International Conference on Data Mining (SDM), 2016 [pdf coming soon]

[10] Learning Structured Output Representation using Deep Conditional Generative Models.
Kihyuk Sohn, Xinchen Yan and Honglak Lee.
In Advances in Neural Information Processing Systems (NIPS), 2015 [pdf][supplementary material][bib]

[9] Improving Object Detection with Deep Convolutional Networks via Bayesian Optimization and Structured Prediction.
Yuting Zhang, Kihyuk Sohn, Ruben Villegas, Gang Pan and Honglak Lee
In Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition (CVPR), 2015 (oral presentation[pdf][supplementary material][tech report][code]
OpenCV People’s Vote Winning Paper [link]

[8] Improved Multimodal Deep Learning with Variation of Information.
Kihyuk Sohn, Wenling Shang and Honglak Lee
In Advances in Neural Information Processing Systems (NIPS), 2014 [pdf][pdf (full)][bib][github]

[7] Learning to Disentangle Factors of Variation with Manifold Interaction.
Scott Reed, Kihyuk Sohn, Yuting Zhang and Honglak Lee
In Proceedings of the 31st International Conference on Machine Learning (ICML), 2014. [pdf][bib][code]

[6] Augmenting CRFs with Boltzmann Machine Shape Priors for Image Labeling.
Kihyuk Sohn*Andrew Kae*, Honglak Lee and Erik Learned-Miller.
In Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition (CVPR), 2013. [pdf][bib][project page][code] (* indicates equal contribution.)

[5] Learning and Selecting Features Jointly with Point-wise Gated Boltzmann Machines.
Kihyuk Sohn, Guanyu Zhou, Chansoo Lee, and Honglak Lee.
In Proceedings of the 30th International Conference on Machine Learning (ICML), 2013. [pdf][bib][supplementary material][project page][code]
(A previous version was presented at the NIPS Workshop on Deep Learning and Unsupervised Feature Learning, 2012.)

[4] Learning Invariant Representations with Local Transformations.
Kihyuk Sohn and Honglak Lee.
In Proceedings of the 29th International Conference on Machine Learning (ICML), 2012. [pdf][bib][github]

[3] Online Incremental Feature Learning with Denoising Autoencoders.
Guanyu Zhou, Kihyuk Sohn, and Honglak Lee.
In Proceedings of the 15th International Conference on Artificial Intelligence and Statistics (AISTATS), JMLR W&CP 22, 2012. [pdf][bib][supplementary material] (oral presentation)
(A previous version was presented at the NIPS Workshop on Deep Learning and Unsupervised Feature Learning, 2011.)

[2] An interpretation of the Cover and Leung capacity region for the MAC with feedback through stochastic control.
Achilleas Anastasopoulos and Kihyuk Sohn.
In Proceedings of IEEE International Conference on Communications (ICC), 2012. [pdf][bib]

[1] Efficient Learning of Sparse, Distributed, Convolutional Feature Representations for Object Recognition.
Kihyuk Sohn, Dae Yon Jung, Honglak Lee, and Alfred Hero III.
In Proceedings of 13th International Conference on Computer Vision (ICCV), 2011. [pdf][bib]



Software