Time-Discounting Convolution for Event Sequences with Ambiguous Timestamps

Takayuki Katsuki, Takayuki Osogami, Akira Koseki, Masaki Ono, Michiharu Kudo, Masaki Makino, Atsushi Suzuki

Research output: Chapter in Book/Report/Conference proceedingConference contribution

Abstract

This paper proposes a method for modeling event sequences with ambiguous timestamps, a time-discounting convolution. Unlike in ordinary time series, time intervals are not constant, small time-shifts have no significant effect, and inputting timestamps or time durations into a model is not effective. The criteria that we require for the modeling are providing robustness against time-shifts or timestamps uncertainty as well as maintaining the essential capabilities of time-series models, i.e., forgetting meaningless past information and handling infinite sequences. The proposed method handles them with a convolutional mechanism across time with specific parameterizations, which efficiently represents the event dependencies in a time-shift invariant manner while discounting the effect of past events, and a dynamic pooling mechanism, which provides robustness against the uncertainty in timestamps and enhances the time-discounting capability by dynamically changing the pooling window size. In our learning algorithm, the decaying and dynamic pooling mechanisms play critical roles in handling infinite and variable length sequences. Numerical experiments on real-world event sequences with ambiguous timestamps and ordinary time series demonstrated the advantages of our method.

Original languageEnglish
Title of host publication2018 IEEE International Conference on Data Mining, ICDM 2018
PublisherInstitute of Electrical and Electronics Engineers Inc.
Pages1085-1090
Number of pages6
ISBN (Electronic)9781538691588
DOIs
Publication statusPublished - 27-12-2018
Event18th IEEE International Conference on Data Mining, ICDM 2018 - Singapore, Singapore
Duration: 17-11-201820-11-2018

Publication series

NameProceedings - IEEE International Conference on Data Mining, ICDM
Volume2018-November
ISSN (Print)1550-4786

Conference

Conference18th IEEE International Conference on Data Mining, ICDM 2018
CountrySingapore
CitySingapore
Period17-11-1820-11-18

Fingerprint

Convolution
Time series
Parameterization
Learning algorithms
Experiments
Uncertainty

All Science Journal Classification (ASJC) codes

  • Engineering(all)

Cite this

Katsuki, T., Osogami, T., Koseki, A., Ono, M., Kudo, M., Makino, M., & Suzuki, A. (2018). Time-Discounting Convolution for Event Sequences with Ambiguous Timestamps. In 2018 IEEE International Conference on Data Mining, ICDM 2018 (pp. 1085-1090). [8594948] (Proceedings - IEEE International Conference on Data Mining, ICDM; Vol. 2018-November). Institute of Electrical and Electronics Engineers Inc.. https://doi.org/10.1109/ICDM.2018.00139
Katsuki, Takayuki ; Osogami, Takayuki ; Koseki, Akira ; Ono, Masaki ; Kudo, Michiharu ; Makino, Masaki ; Suzuki, Atsushi. / Time-Discounting Convolution for Event Sequences with Ambiguous Timestamps. 2018 IEEE International Conference on Data Mining, ICDM 2018. Institute of Electrical and Electronics Engineers Inc., 2018. pp. 1085-1090 (Proceedings - IEEE International Conference on Data Mining, ICDM).
@inproceedings{78787605ee6241558cd9479c4d5a5b78,
title = "Time-Discounting Convolution for Event Sequences with Ambiguous Timestamps",
abstract = "This paper proposes a method for modeling event sequences with ambiguous timestamps, a time-discounting convolution. Unlike in ordinary time series, time intervals are not constant, small time-shifts have no significant effect, and inputting timestamps or time durations into a model is not effective. The criteria that we require for the modeling are providing robustness against time-shifts or timestamps uncertainty as well as maintaining the essential capabilities of time-series models, i.e., forgetting meaningless past information and handling infinite sequences. The proposed method handles them with a convolutional mechanism across time with specific parameterizations, which efficiently represents the event dependencies in a time-shift invariant manner while discounting the effect of past events, and a dynamic pooling mechanism, which provides robustness against the uncertainty in timestamps and enhances the time-discounting capability by dynamically changing the pooling window size. In our learning algorithm, the decaying and dynamic pooling mechanisms play critical roles in handling infinite and variable length sequences. Numerical experiments on real-world event sequences with ambiguous timestamps and ordinary time series demonstrated the advantages of our method.",
author = "Takayuki Katsuki and Takayuki Osogami and Akira Koseki and Masaki Ono and Michiharu Kudo and Masaki Makino and Atsushi Suzuki",
year = "2018",
month = "12",
day = "27",
doi = "10.1109/ICDM.2018.00139",
language = "English",
series = "Proceedings - IEEE International Conference on Data Mining, ICDM",
publisher = "Institute of Electrical and Electronics Engineers Inc.",
pages = "1085--1090",
booktitle = "2018 IEEE International Conference on Data Mining, ICDM 2018",
address = "United States",

}

Katsuki, T, Osogami, T, Koseki, A, Ono, M, Kudo, M, Makino, M & Suzuki, A 2018, Time-Discounting Convolution for Event Sequences with Ambiguous Timestamps. in 2018 IEEE International Conference on Data Mining, ICDM 2018., 8594948, Proceedings - IEEE International Conference on Data Mining, ICDM, vol. 2018-November, Institute of Electrical and Electronics Engineers Inc., pp. 1085-1090, 18th IEEE International Conference on Data Mining, ICDM 2018, Singapore, Singapore, 17-11-18. https://doi.org/10.1109/ICDM.2018.00139

Time-Discounting Convolution for Event Sequences with Ambiguous Timestamps. / Katsuki, Takayuki; Osogami, Takayuki; Koseki, Akira; Ono, Masaki; Kudo, Michiharu; Makino, Masaki; Suzuki, Atsushi.

2018 IEEE International Conference on Data Mining, ICDM 2018. Institute of Electrical and Electronics Engineers Inc., 2018. p. 1085-1090 8594948 (Proceedings - IEEE International Conference on Data Mining, ICDM; Vol. 2018-November).

Research output: Chapter in Book/Report/Conference proceedingConference contribution

TY - GEN

T1 - Time-Discounting Convolution for Event Sequences with Ambiguous Timestamps

AU - Katsuki, Takayuki

AU - Osogami, Takayuki

AU - Koseki, Akira

AU - Ono, Masaki

AU - Kudo, Michiharu

AU - Makino, Masaki

AU - Suzuki, Atsushi

PY - 2018/12/27

Y1 - 2018/12/27

N2 - This paper proposes a method for modeling event sequences with ambiguous timestamps, a time-discounting convolution. Unlike in ordinary time series, time intervals are not constant, small time-shifts have no significant effect, and inputting timestamps or time durations into a model is not effective. The criteria that we require for the modeling are providing robustness against time-shifts or timestamps uncertainty as well as maintaining the essential capabilities of time-series models, i.e., forgetting meaningless past information and handling infinite sequences. The proposed method handles them with a convolutional mechanism across time with specific parameterizations, which efficiently represents the event dependencies in a time-shift invariant manner while discounting the effect of past events, and a dynamic pooling mechanism, which provides robustness against the uncertainty in timestamps and enhances the time-discounting capability by dynamically changing the pooling window size. In our learning algorithm, the decaying and dynamic pooling mechanisms play critical roles in handling infinite and variable length sequences. Numerical experiments on real-world event sequences with ambiguous timestamps and ordinary time series demonstrated the advantages of our method.

AB - This paper proposes a method for modeling event sequences with ambiguous timestamps, a time-discounting convolution. Unlike in ordinary time series, time intervals are not constant, small time-shifts have no significant effect, and inputting timestamps or time durations into a model is not effective. The criteria that we require for the modeling are providing robustness against time-shifts or timestamps uncertainty as well as maintaining the essential capabilities of time-series models, i.e., forgetting meaningless past information and handling infinite sequences. The proposed method handles them with a convolutional mechanism across time with specific parameterizations, which efficiently represents the event dependencies in a time-shift invariant manner while discounting the effect of past events, and a dynamic pooling mechanism, which provides robustness against the uncertainty in timestamps and enhances the time-discounting capability by dynamically changing the pooling window size. In our learning algorithm, the decaying and dynamic pooling mechanisms play critical roles in handling infinite and variable length sequences. Numerical experiments on real-world event sequences with ambiguous timestamps and ordinary time series demonstrated the advantages of our method.

UR - http://www.scopus.com/inward/record.url?scp=85061361168&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=85061361168&partnerID=8YFLogxK

U2 - 10.1109/ICDM.2018.00139

DO - 10.1109/ICDM.2018.00139

M3 - Conference contribution

T3 - Proceedings - IEEE International Conference on Data Mining, ICDM

SP - 1085

EP - 1090

BT - 2018 IEEE International Conference on Data Mining, ICDM 2018

PB - Institute of Electrical and Electronics Engineers Inc.

ER -

Katsuki T, Osogami T, Koseki A, Ono M, Kudo M, Makino M et al. Time-Discounting Convolution for Event Sequences with Ambiguous Timestamps. In 2018 IEEE International Conference on Data Mining, ICDM 2018. Institute of Electrical and Electronics Engineers Inc. 2018. p. 1085-1090. 8594948. (Proceedings - IEEE International Conference on Data Mining, ICDM). https://doi.org/10.1109/ICDM.2018.00139