Incremental dual-memory LSTM in land cover prediction

Xiaowei Jia; Ankush Khandelwal; Guruprasad Nayak; James S Gerber; Kimberly M Carlson; Paul West; Vipin Kumar

doi:10.1145/3097983.3098112

Incremental dual-memory LSTM in land cover prediction

Xiaowei Jia, Ankush Khandelwal, Guruprasad Nayak, James S Gerber, Kimberly M Carlson, Paul West, Vipin Kumar

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

37 Scopus citations

Abstract

Land cover prediction is essential for monitoring global environmental change. Unfortunately, traditional classification models are plagued by temporal variation and emergence of novel/unseen land cover classes in the prediction process. In this paper, we propose an LSTM-based spatiotemporal learning framework with a dual-memory structure. The dual-memory structure captures both long-term and short-term temporal variation patterns, and is updated incrementally to adapt the model to the ever-changing environment. Moreover, we integrate zero-shot learning to identify unseen classes even without labelled samples. Experiments on both synthetic and real-world datasets demonstrate the superiority of the proposed framework over multiple baselines in land cover prediction.

Original language	English (US)
Title of host publication	KDD 2017 - Proceedings of the 23rd ACM SIGKDD International Conference on Knowledge Discovery and Data Mining
Publisher	Association for Computing Machinery
Pages	867-876
Number of pages	10
ISBN (Electronic)	9781450348874
DOIs	https://doi.org/10.1145/3097983.3098112
State	Published - Aug 13 2017
Event	23rd ACM SIGKDD International Conference on Knowledge Discovery and Data Mining, KDD 2017 - Halifax, Canada Duration: Aug 13 2017 → Aug 17 2017

Publication series

Name	Proceedings of the ACM SIGKDD International Conference on Knowledge Discovery and Data Mining
Volume	Part F129685

Other

Other	23rd ACM SIGKDD International Conference on Knowledge Discovery and Data Mining, KDD 2017
Country/Territory	Canada
City	Halifax
Period	8/13/17 → 8/17/17

Bibliographical note

Funding Information:
This work was funded by the NSF Awards 1029711, and Gordon and Betty Moore Foundation and the Belmont Forum/FACCE-JPI funded DEVIL project (NE/M021327/1). Access to [13] computing facilities was provided by NASA Earth Exchange and Minnesota Supercomputing Institute.

Publisher Copyright:
© 2017 ACM.

Keywords

LSTM
Land cover
Zero-short learning

Access

10.1145/3097983.3098112

Cite this

Jia, X., Khandelwal, A., Nayak, G., Gerber, J. S., Carlson, K. M., West, P., & Kumar, V. (2017). Incremental dual-memory LSTM in land cover prediction. In KDD 2017 - Proceedings of the 23rd ACM SIGKDD International Conference on Knowledge Discovery and Data Mining (pp. 867-876). (Proceedings of the ACM SIGKDD International Conference on Knowledge Discovery and Data Mining; Vol. Part F129685). Association for Computing Machinery. https://doi.org/10.1145/3097983.3098112

Incremental dual-memory LSTM in land cover prediction. / Jia, Xiaowei; Khandelwal, Ankush; Nayak, Guruprasad; Gerber, James S; Carlson, Kimberly M; West, Paul ; Kumar, Vipin.

KDD 2017 - Proceedings of the 23rd ACM SIGKDD International Conference on Knowledge Discovery and Data Mining. Association for Computing Machinery, 2017. p. 867-876 (Proceedings of the ACM SIGKDD International Conference on Knowledge Discovery and Data Mining; Vol. Part F129685).

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

Jia, X, Khandelwal, A, Nayak, G, Gerber, JS, Carlson, KM, West, P & Kumar, V 2017, Incremental dual-memory LSTM in land cover prediction. in KDD 2017 - Proceedings of the 23rd ACM SIGKDD International Conference on Knowledge Discovery and Data Mining. Proceedings of the ACM SIGKDD International Conference on Knowledge Discovery and Data Mining, vol. Part F129685, Association for Computing Machinery, pp. 867-876, 23rd ACM SIGKDD International Conference on Knowledge Discovery and Data Mining, KDD 2017, Halifax, Canada, 8/13/17. https://doi.org/10.1145/3097983.3098112

Jia X, Khandelwal A, Nayak G, Gerber JS, Carlson KM, West P et al. Incremental dual-memory LSTM in land cover prediction. In KDD 2017 - Proceedings of the 23rd ACM SIGKDD International Conference on Knowledge Discovery and Data Mining. Association for Computing Machinery. 2017. p. 867-876. (Proceedings of the ACM SIGKDD International Conference on Knowledge Discovery and Data Mining). https://doi.org/10.1145/3097983.3098112

Jia, Xiaowei ; Khandelwal, Ankush ; Nayak, Guruprasad ; Gerber, James S ; Carlson, Kimberly M ; West, Paul ; Kumar, Vipin. / Incremental dual-memory LSTM in land cover prediction. KDD 2017 - Proceedings of the 23rd ACM SIGKDD International Conference on Knowledge Discovery and Data Mining. Association for Computing Machinery, 2017. pp. 867-876 (Proceedings of the ACM SIGKDD International Conference on Knowledge Discovery and Data Mining).

@inproceedings{595a53bcbc474cf0b64a90fa424d8f73,

title = "Incremental dual-memory LSTM in land cover prediction",

abstract = "Land cover prediction is essential for monitoring global environmental change. Unfortunately, traditional classification models are plagued by temporal variation and emergence of novel/unseen land cover classes in the prediction process. In this paper, we propose an LSTM-based spatiotemporal learning framework with a dual-memory structure. The dual-memory structure captures both long-term and short-term temporal variation patterns, and is updated incrementally to adapt the model to the ever-changing environment. Moreover, we integrate zero-shot learning to identify unseen classes even without labelled samples. Experiments on both synthetic and real-world datasets demonstrate the superiority of the proposed framework over multiple baselines in land cover prediction.",

keywords = "LSTM, Land cover, Zero-short learning",

author = "Xiaowei Jia and Ankush Khandelwal and Guruprasad Nayak and Gerber, {James S} and Carlson, {Kimberly M} and Paul West and Vipin Kumar",

note = "Funding Information: This work was funded by the NSF Awards 1029711, and Gordon and Betty Moore Foundation and the Belmont Forum/FACCE-JPI funded DEVIL project (NE/M021327/1). Access to [13] computing facilities was provided by NASA Earth Exchange and Minnesota Supercomputing Institute. Publisher Copyright: {\textcopyright} 2017 ACM.; 23rd ACM SIGKDD International Conference on Knowledge Discovery and Data Mining, KDD 2017 ; Conference date: 13-08-2017 Through 17-08-2017",

year = "2017",

month = aug,

day = "13",

doi = "10.1145/3097983.3098112",

language = "English (US)",

series = "Proceedings of the ACM SIGKDD International Conference on Knowledge Discovery and Data Mining",

publisher = "Association for Computing Machinery",

pages = "867--876",

booktitle = "KDD 2017 - Proceedings of the 23rd ACM SIGKDD International Conference on Knowledge Discovery and Data Mining",

}

TY - GEN

T1 - Incremental dual-memory LSTM in land cover prediction

AU - Jia, Xiaowei

AU - Khandelwal, Ankush

AU - Nayak, Guruprasad

AU - Gerber, James S

AU - Carlson, Kimberly M

AU - West, Paul

AU - Kumar, Vipin

N1 - Funding Information: This work was funded by the NSF Awards 1029711, and Gordon and Betty Moore Foundation and the Belmont Forum/FACCE-JPI funded DEVIL project (NE/M021327/1). Access to [13] computing facilities was provided by NASA Earth Exchange and Minnesota Supercomputing Institute. Publisher Copyright: © 2017 ACM.

PY - 2017/8/13

Y1 - 2017/8/13

N2 - Land cover prediction is essential for monitoring global environmental change. Unfortunately, traditional classification models are plagued by temporal variation and emergence of novel/unseen land cover classes in the prediction process. In this paper, we propose an LSTM-based spatiotemporal learning framework with a dual-memory structure. The dual-memory structure captures both long-term and short-term temporal variation patterns, and is updated incrementally to adapt the model to the ever-changing environment. Moreover, we integrate zero-shot learning to identify unseen classes even without labelled samples. Experiments on both synthetic and real-world datasets demonstrate the superiority of the proposed framework over multiple baselines in land cover prediction.

AB - Land cover prediction is essential for monitoring global environmental change. Unfortunately, traditional classification models are plagued by temporal variation and emergence of novel/unseen land cover classes in the prediction process. In this paper, we propose an LSTM-based spatiotemporal learning framework with a dual-memory structure. The dual-memory structure captures both long-term and short-term temporal variation patterns, and is updated incrementally to adapt the model to the ever-changing environment. Moreover, we integrate zero-shot learning to identify unseen classes even without labelled samples. Experiments on both synthetic and real-world datasets demonstrate the superiority of the proposed framework over multiple baselines in land cover prediction.

KW - LSTM

KW - Land cover

KW - Zero-short learning

UR - http://www.scopus.com/inward/record.url?scp=85029035936&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=85029035936&partnerID=8YFLogxK

U2 - 10.1145/3097983.3098112

DO - 10.1145/3097983.3098112

M3 - Conference contribution

AN - SCOPUS:85029035936

T3 - Proceedings of the ACM SIGKDD International Conference on Knowledge Discovery and Data Mining

SP - 867

EP - 876

BT - KDD 2017 - Proceedings of the 23rd ACM SIGKDD International Conference on Knowledge Discovery and Data Mining

PB - Association for Computing Machinery

T2 - 23rd ACM SIGKDD International Conference on Knowledge Discovery and Data Mining, KDD 2017

Y2 - 13 August 2017 through 17 August 2017

ER -

Incremental dual-memory LSTM in land cover prediction

Abstract

Publication series

Other

Bibliographical note

Keywords

Access

Other files and links

Fingerprint

Cite this