TY - GEN
T1 - The role of sparsity and dynamics in extracting information sparsely encoded in very large data sets
AU - Camps, O.
AU - Lagoa, C.
AU - Sznaier, M.
N1 - Funding Information:
M. Sznaier and O. Camps are with the ECE Dept., Northeastern University, Boston, MA, 02115, email: {msznaier,camps}@coe.neu.edu. C. Lagoa is with the EE Dept., Penn State University, University Park, PA, 16802, email: [email protected]. This work was supported in part by NSF grants ECCS–1201973, IIS–1318145, CNS–1329422 and ECCS–1404163; AFOSR grant FA9550-12-1-0271 and the Alert DHS Center of Excellence under Award Number 2008-ST-061-ED0001.
Publisher Copyright:
© 2016 Institute of Electrical and Electronics Engineers Inc.. All rights reserved.
PY - 2016/10/10
Y1 - 2016/10/10
N2 - The past few years have seen an exponential growth in data collection capabilities. Unfortunately, the ability to process this vast amount of data has not kept pace with this growth. Taking full advantage of these increased capabilities requires scalable, computationally efficient algorithms to timely and robustly extract actionable information from the very large data sets generated by the sensors. The goal of this tutorial paper is to illustrate the central role that tools originally developed in the context of systems theory, can play in accomplishing this task. Specifically, we show that many of these problems can be recast into an identification form that exhibit a sparse underlying structure. In turn, this sparsity can be exploited to recast the problem into a convex optimization form that can be efficiently solved with first order methods.
AB - The past few years have seen an exponential growth in data collection capabilities. Unfortunately, the ability to process this vast amount of data has not kept pace with this growth. Taking full advantage of these increased capabilities requires scalable, computationally efficient algorithms to timely and robustly extract actionable information from the very large data sets generated by the sensors. The goal of this tutorial paper is to illustrate the central role that tools originally developed in the context of systems theory, can play in accomplishing this task. Specifically, we show that many of these problems can be recast into an identification form that exhibit a sparse underlying structure. In turn, this sparsity can be exploited to recast the problem into a convex optimization form that can be efficiently solved with first order methods.
UR - http://www.scopus.com/inward/record.url?scp=85113870333&partnerID=8YFLogxK
UR - http://www.scopus.com/inward/citedby.url?scp=85113870333&partnerID=8YFLogxK
U2 - 10.1109/CCA.2016.7587864
DO - 10.1109/CCA.2016.7587864
M3 - Conference contribution
AN - SCOPUS:85113870333
T3 - 2016 IEEE Conference on Control Applications, CCA 2016
SP - 398
EP - 409
BT - 2016 IEEE Conference on Control Applications, CCA 2016
PB - Institute of Electrical and Electronics Engineers Inc.
T2 - 2016 IEEE Conference on Control Applications, CCA 2016
Y2 - 19 September 2016 through 22 September 2016
ER -