Publication

Recursive Training of 2D-3D Convolutional Networks for Neuronal Boundary Detection.

Kisuk Lee, Aleksandar Zlateski, Ashwin Vishwanathan, & H. Sebastian Seung. In NIPS 2015.

overview_only.png

Space-Time Wiring Specificity Supports Direction Selectivity in the Retina

Jinseop S. Kim, Matthew J. Greene, Aleksandar Zlateski, Kisuk Lee, Mark Richardson, Srinivas C. Turaga, Michael Purcaro, Matthew Balkam, Amy Robinson, Bardia F. Behabadi, Michael Campos, Winfried Denk, H. Sebastian Seung & the EyeWirers. Nature 509, pp.331-336, 2014.

Kim2014_fig.png

In this paper, I made a modest contribution by developing a system for objectively measuring accuracy of the players (a.k.a. EyeWirers) in EyeWire [1], the world's first game to map the brain that is designed to speed up the process of analyzing EM images of neural tissue by combining machine and crowd intelligence. Based on the accuracy system I developed, I analyzed and compared the accuracy between individual EyeWirers, EyeWirer consensus, and artificial intelligence (AI), and found that EyeWirer consensus was much more accurate than any individual EyeWirer or the AI. I also found an important trend that EyeWirer accuracy improves over the course of tens of hours of practice.

Cooking Action Recognition via Spatio-Temporal Feature Learning based on ISA (Extended, in Korean)

Kisuk Lee, Eun-Sol Kim, Karinne Ramirez Amaro, Michael Beetz & Byoung-Tak Zhang. Journal of the Korean Institute of Information Science Society: Computing Practices and Letters 19(8), pp.434-438, 2013.


Human Cooking Action Recognition via Spatio-Temporal Feature Learning based on ISA (in Korean)

Kisuk Lee, Eun-Sol Kim, Karinne Ramirez Amaro, Michael Beetz & Byoung-Tak Zhang. In Proceedings of the Korea Computing Congress (KCC 2012) Vol.39 2(B), pp.183-185, Nov. 2012.


interest_point.jpg

In this paper, we adopted a convolutional stacked independent subspace analysis (ISA) algorithm [2] to learn spatio-temporal features for human cooking action recognition. This work was presented at the Korea Computing Congress (KCC2012) in November 2012, and selected as one of the best oral presentation papers, leading to an invited publication in the Journal of the Korean Institute of Information Science Scoiety: Computing Practices and Letters.

Searching for Spatio-Temopral Pattern in EEG Signal with Hypernetwork (in Korean)

Eun-Sol Kim, Chung-Yeon Lee, Kisuk Lee, Hyunmin Lee, JoonShik Kim & Byoung-Tak Zhang. In Proceedings of the Korea Computing Congress (KCC 2011) Vol.38 1(C), pp.331-334, Nov. 2011.

hypernet_EEG.png

In this project, we searched for spatio-temporally dominant patterns in EEG signals with Hypernetwork [3], a random hypergraph structure representing higher-order probabilistic relations of data by an evolutionary self-organizing process based on molecular self-assembly. Specifically, we developed a novel Hypernetwork architecture incorporating an additional temporal dimension to detect dominant spatio-temporal patterns in EEG signals. This work was presented at the Korea Computing Congress (KCC2011) in June 2011.

Bibliography
2. Le, Q.V., Zou, W.Y., Yeung, S.Y. & Ng, A.Y. 2011. Learning Hierarchical Invariant Spatio-Temporal Features for Action Recognition with Independent Subspace Analysis. In CPVR 2011.
3. Zhang, B.-T. 2008. Hypernetworks: A Molecular Evolutionary Architecture for Cognitive Learning and Memory. IEEE Computational Intelligence Magazine 3(3), pp.49-63.
Unless otherwise stated, the content of this page is licensed under Creative Commons Attribution-ShareAlike 3.0 License