BLPA

Bayesian Learn-Predict-Adjust Method for Online Detection of Recurrent Changepoints

Alexandr Maslov, Mykola Pechenizkiy, Yulong Pei, Indre Zliobaite, Alexander Shklyaev, Tommi Kärkkäinen, Jaakko Hollmén

Forskningsoutput: Kapitel i bok/rapport/konferenshandlingKonferensbidragVetenskapligPeer review

Sammanfattning

Online changepoint detection is an important task for machine learning in changing environments, as it signals when the learning model needs to be updated. Presence of noise that can be mistaken for real changes makes it difficult to develop an effective approach that would have a low false alarm rate and being able to detect all the changes with a minimal delay. In this paper we study how performance of popular Bayesian online detectors can be improved in case of recurrent changes. Modelling recurrence allows us to anticipate future changepoints and predict their locations in time. We propose an approach for inducing and integrating recurrence information in the streaming settings, and demonstrate its effectiveness on synthetic and real-world human activity datasets.
Originalspråkengelska
Titel på gästpublikationIJCNN 2017 : The International Joint Conference on Neural Networks
Antal sidor6
UtgivningsortPiscataway, NJ
FörlagInstitute of Electrical and Electronics Engineers
Utgivningsdatum2017
Sidor1916-1923
ISBN (elektroniskt)978-1-5090-6181-5
DOI
StatusPublicerad - 2017
MoE-publikationstypA4 Artikel i en konferenspublikation
EvenemangInternational Joint Conferenceon Neural Networks - Anchorage, Förenta Staterna (USA)
Varaktighet: 14 maj 201719 maj 2017
Konferensnummer: 30

Publikationsserier

NamnProceedings of ... International Joint Conference on Neural Networks
FörlagInstitute of Electrical and Electronic Engineers
ISSN (elektroniskt)2161-4407

Vetenskapsgrenar

  • 1171 Geovetenskaper

Citera det här

Maslov, A., Pechenizkiy, M., Pei, Y., Zliobaite, I., Shklyaev, A., Kärkkäinen, T., & Hollmén, J. (2017). BLPA: Bayesian Learn-Predict-Adjust Method for Online Detection of Recurrent Changepoints. I IJCNN 2017: The International Joint Conference on Neural Networks (s. 1916-1923). (Proceedings of ... International Joint Conference on Neural Networks ). Piscataway, NJ : Institute of Electrical and Electronics Engineers. https://doi.org/10.1109/IJCNN.2017.7966085
Maslov, Alexandr ; Pechenizkiy, Mykola ; Pei, Yulong ; Zliobaite, Indre ; Shklyaev, Alexander ; Kärkkäinen, Tommi ; Hollmén, Jaakko. / BLPA : Bayesian Learn-Predict-Adjust Method for Online Detection of Recurrent Changepoints. IJCNN 2017: The International Joint Conference on Neural Networks. Piscataway, NJ : Institute of Electrical and Electronics Engineers, 2017. s. 1916-1923 (Proceedings of ... International Joint Conference on Neural Networks ).
@inproceedings{afb1f0737c7f4891ae8ea9735ee7ab44,
title = "BLPA: Bayesian Learn-Predict-Adjust Method for Online Detection of Recurrent Changepoints",
abstract = "Online changepoint detection is an important task for machine learning in changing environments, as it signals when the learning model needs to be updated. Presence of noise that can be mistaken for real changes makes it difficult to develop an effective approach that would have a low false alarm rate and being able to detect all the changes with a minimal delay. In this paper we study how performance of popular Bayesian online detectors can be improved in case of recurrent changes. Modelling recurrence allows us to anticipate future changepoints and predict their locations in time. We propose an approach for inducing and integrating recurrence information in the streaming settings, and demonstrate its effectiveness on synthetic and real-world human activity datasets.",
keywords = "1171 Geosciences",
author = "Alexandr Maslov and Mykola Pechenizkiy and Yulong Pei and Indre Zliobaite and Alexander Shklyaev and Tommi K{\"a}rkk{\"a}inen and Jaakko Hollm{\'e}n",
note = "Volume: Proceeding volume:",
year = "2017",
doi = "10.1109/IJCNN.2017.7966085",
language = "English",
series = "Proceedings of ... International Joint Conference on Neural Networks",
publisher = "Institute of Electrical and Electronics Engineers",
pages = "1916--1923",
booktitle = "IJCNN 2017",
address = "United States",

}

Maslov, A, Pechenizkiy, M, Pei, Y, Zliobaite, I, Shklyaev, A, Kärkkäinen, T & Hollmén, J 2017, BLPA: Bayesian Learn-Predict-Adjust Method for Online Detection of Recurrent Changepoints. i IJCNN 2017: The International Joint Conference on Neural Networks. Proceedings of ... International Joint Conference on Neural Networks , Institute of Electrical and Electronics Engineers, Piscataway, NJ , s. 1916-1923, International Joint Conferenceon Neural Networks, Anchorage, Förenta Staterna (USA), 14/05/2017. https://doi.org/10.1109/IJCNN.2017.7966085

BLPA : Bayesian Learn-Predict-Adjust Method for Online Detection of Recurrent Changepoints. / Maslov, Alexandr; Pechenizkiy, Mykola; Pei, Yulong; Zliobaite, Indre; Shklyaev, Alexander; Kärkkäinen, Tommi; Hollmén, Jaakko.

IJCNN 2017: The International Joint Conference on Neural Networks. Piscataway, NJ : Institute of Electrical and Electronics Engineers, 2017. s. 1916-1923 (Proceedings of ... International Joint Conference on Neural Networks ).

Forskningsoutput: Kapitel i bok/rapport/konferenshandlingKonferensbidragVetenskapligPeer review

TY - GEN

T1 - BLPA

T2 - Bayesian Learn-Predict-Adjust Method for Online Detection of Recurrent Changepoints

AU - Maslov, Alexandr

AU - Pechenizkiy, Mykola

AU - Pei, Yulong

AU - Zliobaite, Indre

AU - Shklyaev, Alexander

AU - Kärkkäinen, Tommi

AU - Hollmén, Jaakko

N1 - Volume: Proceeding volume:

PY - 2017

Y1 - 2017

N2 - Online changepoint detection is an important task for machine learning in changing environments, as it signals when the learning model needs to be updated. Presence of noise that can be mistaken for real changes makes it difficult to develop an effective approach that would have a low false alarm rate and being able to detect all the changes with a minimal delay. In this paper we study how performance of popular Bayesian online detectors can be improved in case of recurrent changes. Modelling recurrence allows us to anticipate future changepoints and predict their locations in time. We propose an approach for inducing and integrating recurrence information in the streaming settings, and demonstrate its effectiveness on synthetic and real-world human activity datasets.

AB - Online changepoint detection is an important task for machine learning in changing environments, as it signals when the learning model needs to be updated. Presence of noise that can be mistaken for real changes makes it difficult to develop an effective approach that would have a low false alarm rate and being able to detect all the changes with a minimal delay. In this paper we study how performance of popular Bayesian online detectors can be improved in case of recurrent changes. Modelling recurrence allows us to anticipate future changepoints and predict their locations in time. We propose an approach for inducing and integrating recurrence information in the streaming settings, and demonstrate its effectiveness on synthetic and real-world human activity datasets.

KW - 1171 Geosciences

U2 - 10.1109/IJCNN.2017.7966085

DO - 10.1109/IJCNN.2017.7966085

M3 - Conference contribution

T3 - Proceedings of ... International Joint Conference on Neural Networks

SP - 1916

EP - 1923

BT - IJCNN 2017

PB - Institute of Electrical and Electronics Engineers

CY - Piscataway, NJ

ER -

Maslov A, Pechenizkiy M, Pei Y, Zliobaite I, Shklyaev A, Kärkkäinen T et al. BLPA: Bayesian Learn-Predict-Adjust Method for Online Detection of Recurrent Changepoints. I IJCNN 2017: The International Joint Conference on Neural Networks. Piscataway, NJ : Institute of Electrical and Electronics Engineers. 2017. s. 1916-1923. (Proceedings of ... International Joint Conference on Neural Networks ). https://doi.org/10.1109/IJCNN.2017.7966085