國立虎尾科技大學 |

Identification and approximation of the structure of networks of stochastic processes.

紀錄類型:	書目-語言資料,手稿 : Monograph/item
正題名/作者:	Identification and approximation of the structure of networks of stochastic processes./
作者:	Quinn, Christopher John.
面頁冊數:	1 online resource (121 pages)
附註:	Source: Dissertation Abstracts International, Volume: 76-06(E), Section: B.
標題:	Electrical engineering. -
電子資源:	click for full text (PQDT)
ISBN:	9781321514520

Identification and approximation of the structure of networks of stochastic processes.
Quinn, Christopher John.

Identification and approximation of the structure of networks of stochastic processes. - 1 online resource (121 pages)

Source: Dissertation Abstracts International, Volume: 76-06(E), Section: B.

Thesis (Ph.D.)--University of Illinois at Urbana-Champaign, 2014.

Includes bibliographical references

We propose a framework to infer influences between agents in a network using only observed time series. The framework is general---it does not require any particular class of models for the dynamics. It includes graphical models to depict influences in the network, algorithms to identify and approximate the graphs, and techniques to estimate directed information, an information theoretic quantity that measures influence, from data. We demonstrate the utility of the methods by identifying influences between neurons in a primate as well as between news agencies and users in the Twitter network.

Electronic reproduction.
Ann Arbor, Mich. :
ProQuest,
2018

Mode of access: World Wide Web

ISBN: 9781321514520Subjects--Topical Terms:

596380
Electrical engineering.
Index Terms--Genre/Form:

554714
Electronic books.

Identification and approximation of the structure of networks of stochastic processes.
LDR:04490ntm a2200397K 4500 001 913760
005 20180622095237.5
006 m o u
007 cr mn||||a|a||
008 190606s2014 xx obm 000 0 eng d
020 $a 9781321514520
035 $a (MiAaPQ)AAI3673827
035 $a (MiAaPQ)50685
035 $a AAI3673827
040 $a MiAaPQ $b eng $c MiAaPQ
100 1 $a Quinn, Christopher John. $3 1186727
245 1 0 $a Identification and approximation of the structure of networks of stochastic processes.
264 0 $c 2014
300 $a 1 online resource (121 pages)
336 $a text $b txt $2 rdacontent
337 $a computer $b c $2 rdamedia
338 $a online resource $b cr $2 rdacarrier
500 $a Source: Dissertation Abstracts International, Volume: 76-06(E), Section: B.
500 $a Adviser: Negar Kiyavash.
502 $a Thesis (Ph.D.)--University of Illinois at Urbana-Champaign, 2014.
504 $a Includes bibliographical references
520 $a We propose a framework to infer influences between agents in a network using only observed time series. The framework is general---it does not require any particular class of models for the dynamics. It includes graphical models to depict influences in the network, algorithms to identify and approximate the graphs, and techniques to estimate directed information, an information theoretic quantity that measures influence, from data. We demonstrate the utility of the methods by identifying influences between neurons in a primate as well as between news agencies and users in the Twitter network.
520 $a We introduce two graphical models to concisely represent causal influences between agents in a network. The first, the minimal generative model graph, reflects a minimal state space description of relationships. The second, the directed information graph, is a statistical approach similar to conventional graphical models and uses directed information to generalize Granger causality. Although they are motivated differently, we show that under minimal assumptions, the graphs are equivalent.
520 $a In order to identify the underlying graph, we present several algorithms. In general, joint statistics of the whole network are needed. An algorithm that uses the minimal-dimension statistics necessary when upper bounds on the in-degrees are known is presented. In the event that the upper bounds are not valid, the result is nonetheless an optimal approximation. An adaptive algorithm is introduced that uses near minimal-dimension statistics but does not require assumptions on the in-degree bound.
520 $a Several algorithms to optimally approximate the graph are proposed. The quality of an approximation is measured by Kullback-Leibler divergence between the full joint distribution and the distribution induced by the approximation. The first class of approximations are directed trees. We then discuss algorithms to find the best connected and unconstrained approximations that have user-specified in-degrees to incorporate more dynamics. A greedy search algorithm is shown to identify near-optimal approximations of these classes.
520 $a The algorithms require calculations of directed information. For the setting when directed information is estimated from data, we characterize the sample-complexity of two plug-in directed information estimators. Their performance is similar to standard results for statistical estimation with i.i.d. data. When point estimates of directed information are not reliable, we compute confidence intervals. Furthermore, we propose algorithms that use confidence intervals to identify graph approximations that are robust to estimation error. We also propose a consistent parametric estimation technique analogous to the asymptotic equipartition property.
520 $a Last, we demonstrate the effectiveness of the proposed algorithms through simulations and data analysis. The methods identify influences between neurons in a primate which give rise to observed regional information transfer observed by a collaborator. The framework also identifies which news agencies influence which users in the Twitter network by analyzing only tweet times. The algorithms determine influences with high precision.
533 $a Electronic reproduction. $b Ann Arbor, Mich. : $c ProQuest, $d 2018
538 $a Mode of access: World Wide Web
650 4 $a Electrical engineering. $3 596380
650 4 $a Computer science. $3 573171
650 4 $a Statistics. $3 556824
655 7 $a Electronic books. $2 local $3 554714
690 $a 0544
690 $a 0984
690 $a 0463
710 2 $a ProQuest Information and Learning Co. $3 1178819
710 2 $a University of Illinois at Urbana-Champaign. $b Electrical and Computer Engineering. $3 1186728
856 4 0 $u http://pqdd.sinica.edu.tw/twdaoapp/servlet/advanced?query=3673827 $z click for full text (PQDT)