To prospective students:
I am looking for multiple Ph.D. students to join my research group in 2015. Those who are interested in large-scale data storage systems, coding for data storage, information theory and communications are welcome to contact me through email.
Department of EECS, The University of Tennessee at Knoxville / Aug. 2014 -
Member of Technical Staff-Research
AT&T Labs-Research, Bedminster, NJ / Oct. 2007 - Jul. 2014
EPFL, Lausanne, Switzerland / Sep. 2005 - Sep. 2007
Cornell University, Ithaca, NY
Ph.D. Electrical and Computer Engineering / Aug. 2005
Tsinghua University, Beijing, China
Bachelor of Engineering (BE), Electronic Engineering / Jul. 2000
AT&T Key Contributor Award
For technical contribution in AT&T / 2010, 2011, 2013
Liu-Memorial Award, Cornell University
For excellence in graduate study and research / 2004
Characterizing the rate-region of the (4,3,3) exact-repair regenerating codes
C. Tian, IEEE JSAC on Comm. Methodologies for the Next-Gen Storage Systems, May 2014.
A fundamental problem in distributed data storage systems is whether the capacity of the functional-repair regenerating codes is the same as the exact-repair version. This paper answers in the negative through a novel computer-aided approach.
Optimality and approximate optimality of source-channel separation in networks
C. Tian, J. Chen, S. N. Diggavi and S. Shamai, IEEE Trans. Inform. Theory, Feb. 2014
Underlying the layered architecture of most communication networks (e.g., Internet) is the optimistic assumption that such a separation does not incur any loss, which is however not generally true in many multiuser scenarios. Fortunately, here we show for several general classes of communication networks, separation is either optimal or close to optimal.
Accelerated bilateral filtering with block skipping
C. Tian, and S. Krishnan, IEEE Signal Processing Letters, May 2013
An improvement is proposed which provides up to 5x speedup for the fastest bilateral filtering algorithm in the literature.
The achievable distortion region of sending a bivariate Gaussian source on the Gaussian broadcast channel
C. Tian, S. N. Diggavi and S. Shamai, IEEE Trans. Inform. Theory, Oct. 2011
A complete distortion region characterization for the joint source-channel coding problem of broadcasting bivariate Gaussians. This is the first case where a hybrid signaling scheme is found to be optimal, while neither digital signaling nor analog signaling is sufficient.
Approximating the Gaussian multiple description rate region under symmetric distortion constraints
C. Tian, S. Mohajer and S. N. Diggavi, IEEE Trans. Inform. Theory, Aug. 2009
An approximate solution for a long-standing open problem in lossy distributed data storage. The surprise here is that a simple combination of scalable coding and unequal loss protection is in fact approximately optimal.
Quantcast QFS is designed to better accommodate the map-reduce framework by taking into account of the more recent hardware architecture. It originally supports on up to 3 parities in erasure code setting. In this project, we add more flexible erasure codes to allow more than 3 parities.
Implmentation of a set of regenerating codes and locally repairable codes for distributed data storage.