On Information Bottleneck for Gaussian Processes

Michael Dikshtein; Nir Weinberger; Shlomo Shamai Shitz

doi:10.1109/ITW54588.2022.9965939

On Information Bottleneck for Gaussian Processes

Michael Dikshtein, Nir Weinberger, Shlomo Shamai Shitz

Research output: Chapter in Book/Report/Conference proceeding › Conference contribution › peer-review

Abstract

The information bottleneck (IB) problem of jointly stationary Gaussian sources is considered. A water-filling solution for the IB rate is given in terms of its SNR spectrum and whose rate is attained via frequency domain test-channel realization. A time-domain realization of the IB rate, based on linear prediction, is also proposed, which lends itself to an efficient implementation of the corresponding remote source-coding problem. A compound version of the problem is addressed, in which the joint distribution of the source is not precisely specified but rather in terms of a lower bound on the guaranteed mutual information. It is proved that a white SNR spectrum is optimal for this setting.

Original language	English
Title of host publication	2022 IEEE Information Theory Workshop, ITW 2022
Publisher	Institute of Electrical and Electronics Engineers Inc.
Pages	546-551
Number of pages	6
ISBN (Electronic)	9781665483414
DOIs	https://doi.org/10.1109/ITW54588.2022.9965939
State	Published - 2022
Externally published	Yes
Event	2022 IEEE Information Theory Workshop, ITW 2022 - Mumbai, India Duration: 1 Nov 2022 → 9 Nov 2022

Publication series

Name	2022 IEEE Information Theory Workshop, ITW 2022

Conference

Conference	2022 IEEE Information Theory Workshop, ITW 2022
Country/Territory	India
City	Mumbai
Period	1/11/22 → 9/11/22

Access to Document

10.1109/ITW54588.2022.9965939

Cite this

@inproceedings{14a241746c4b44d28c62d5c1487fd50c,

title = "On Information Bottleneck for Gaussian Processes",

abstract = "The information bottleneck (IB) problem of jointly stationary Gaussian sources is considered. A water-filling solution for the IB rate is given in terms of its SNR spectrum and whose rate is attained via frequency domain test-channel realization. A time-domain realization of the IB rate, based on linear prediction, is also proposed, which lends itself to an efficient implementation of the corresponding remote source-coding problem. A compound version of the problem is addressed, in which the joint distribution of the source is not precisely specified but rather in terms of a lower bound on the guaranteed mutual information. It is proved that a white SNR spectrum is optimal for this setting.",

author = "Michael Dikshtein and Nir Weinberger and {Shamai Shitz}, Shlomo",

note = "Publisher Copyright: {\textcopyright} 2022 IEEE.; 2022 IEEE Information Theory Workshop, ITW 2022 ; Conference date: 01-11-2022 Through 09-11-2022",

year = "2022",

doi = "10.1109/ITW54588.2022.9965939",

language = "אנגלית",

series = "2022 IEEE Information Theory Workshop, ITW 2022",

publisher = "Institute of Electrical and Electronics Engineers Inc.",

pages = "546--551",

booktitle = "2022 IEEE Information Theory Workshop, ITW 2022",

address = "ארצות הברית",

}

Dikshtein, M, Weinberger, N & Shamai Shitz, S 2022, On Information Bottleneck for Gaussian Processes. in 2022 IEEE Information Theory Workshop, ITW 2022. 2022 IEEE Information Theory Workshop, ITW 2022, Institute of Electrical and Electronics Engineers Inc., pp. 546-551, 2022 IEEE Information Theory Workshop, ITW 2022, Mumbai, India, 1/11/22. https://doi.org/10.1109/ITW54588.2022.9965939

On Information Bottleneck for Gaussian Processes. / Dikshtein, Michael; Weinberger, Nir; Shamai Shitz, Shlomo.
2022 IEEE Information Theory Workshop, ITW 2022. Institute of Electrical and Electronics Engineers Inc., 2022. p. 546-551 (2022 IEEE Information Theory Workshop, ITW 2022).

Research output: Chapter in Book/Report/Conference proceeding › Conference contribution › peer-review

TY - GEN

T1 - On Information Bottleneck for Gaussian Processes

AU - Dikshtein, Michael

AU - Weinberger, Nir

AU - Shamai Shitz, Shlomo

PY - 2022

Y1 - 2022

N2 - The information bottleneck (IB) problem of jointly stationary Gaussian sources is considered. A water-filling solution for the IB rate is given in terms of its SNR spectrum and whose rate is attained via frequency domain test-channel realization. A time-domain realization of the IB rate, based on linear prediction, is also proposed, which lends itself to an efficient implementation of the corresponding remote source-coding problem. A compound version of the problem is addressed, in which the joint distribution of the source is not precisely specified but rather in terms of a lower bound on the guaranteed mutual information. It is proved that a white SNR spectrum is optimal for this setting.

AB - The information bottleneck (IB) problem of jointly stationary Gaussian sources is considered. A water-filling solution for the IB rate is given in terms of its SNR spectrum and whose rate is attained via frequency domain test-channel realization. A time-domain realization of the IB rate, based on linear prediction, is also proposed, which lends itself to an efficient implementation of the corresponding remote source-coding problem. A compound version of the problem is addressed, in which the joint distribution of the source is not precisely specified but rather in terms of a lower bound on the guaranteed mutual information. It is proved that a white SNR spectrum is optimal for this setting.

UR - http://www.scopus.com/inward/record.url?scp=85144595095&partnerID=8YFLogxK

U2 - 10.1109/ITW54588.2022.9965939

DO - 10.1109/ITW54588.2022.9965939

M3 - ???researchoutput.researchoutputtypes.contributiontobookanthology.conference???

AN - SCOPUS:85144595095

T3 - 2022 IEEE Information Theory Workshop, ITW 2022

SP - 546

EP - 551

BT - 2022 IEEE Information Theory Workshop, ITW 2022

PB - Institute of Electrical and Electronics Engineers Inc.

T2 - 2022 IEEE Information Theory Workshop, ITW 2022

Y2 - 1 November 2022 through 9 November 2022

ER -

On Information Bottleneck for Gaussian Processes

Abstract

Publication series

Conference

Access to Document

Other files and links

Fingerprint

Cite this