EVOLUTIONARY OPTIMIZATION OF NEURAL ARCHITECTURES IN REMOTE SENSING CLASSIFICATION PROBLEMS

Daniel Coquelin; Rocco Sedona; Morris Riedel; Markus Götz

doi:10.1109/IGARSS47720.2021.9554309

EVOLUTIONARY OPTIMIZATION OF NEURAL ARCHITECTURES IN REMOTE SENSING CLASSIFICATION PROBLEMS

Daniel Coquelin
, Rocco Sedona
, Morris Riedel
, Markus Götz

Iðnaðarverkfræði-, vélaverkfræði- og tölvunarfræðideild

Háskóli Íslands

Rannsóknarafurð: Framlag á ráðstefnu › Vísindagrein › ritrýni

Útdráttur

BigEarthNet is one of the standard large remote sensing datasets. It has been shown previously that neural networks are effective tools to classify the image patches in this data. However, finding the optimum network hyperparameters and architecture to accurately classify the image patches in BigEarthNet remains a challenge. Searching for more accurate models manually is extremely time consuming and labour intensive. Hence, a systematic approach is advisable. One possibility is automated evolutionary Neural Architecture Search (NAS). With this NAS many of the commonly used network hyperparameters, such as loss functions, are eliminated and a more accurate network is determined.

Upprunalegt tungumál	Enska
Síður	1587-1590
Síðufjöldi	4
DOI	https://doi.org/10.1109/IGARSS47720.2021.9554309
Útgáfustaða	Útgefið - 2021
Viðburður	2021 IEEE International Geoscience and Remote Sensing Symposium, IGARSS 2021 - Brussels, Belgía Tímalengd: 12 júl. 2021 → 16 júl. 2021

Ráðstefna

Ráðstefna	2021 IEEE International Geoscience and Remote Sensing Symposium, IGARSS 2021
Land/Yfirráðasvæði	Belgía
Borg/bær	Brussels
Tímabil	12/07/21 → 16/07/21

Athugasemd

Aðgangur að skjali

10.1109/IGARSS47720.2021.9554309

Vitna í þetta

@conference{dfca517a070d40749e2c68490c4f4dfc,

title = "EVOLUTIONARY OPTIMIZATION OF NEURAL ARCHITECTURES IN REMOTE SENSING CLASSIFICATION PROBLEMS",

abstract = "BigEarthNet is one of the standard large remote sensing datasets. It has been shown previously that neural networks are effective tools to classify the image patches in this data. However, finding the optimum network hyperparameters and architecture to accurately classify the image patches in BigEarthNet remains a challenge. Searching for more accurate models manually is extremely time consuming and labour intensive. Hence, a systematic approach is advisable. One possibility is automated evolutionary Neural Architecture Search (NAS). With this NAS many of the commonly used network hyperparameters, such as loss functions, are eliminated and a more accurate network is determined.",

keywords = "Classification, Evolutionary Algorithms, NAS, Neural Architecture Search, Remote Sensing",

author = "Daniel Coquelin and Rocco Sedona and Morris Riedel and Markus G{\"o}tz",

note = "Funding Information: This work was performed on the supercomputer ForHLR funded by the Ministry of Science, Research and the Arts Baden-W{\"u}rttemberg and by the Federal Ministry of Education and Research. This work is supported by the Helmholtz Association{\textquoteright}s Initiative and Networking Funds under under the Helmholtz AI platform grant and the Helmholtz Analytics Framework collaboration. Publisher Copyright: {\textcopyright} 2021 IEEE; 2021 IEEE International Geoscience and Remote Sensing Symposium, IGARSS 2021 ; Conference date: 12-07-2021 Through 16-07-2021",

year = "2021",

doi = "10.1109/IGARSS47720.2021.9554309",

language = "English",

pages = "1587--1590",

}

TY - CONF

T1 - EVOLUTIONARY OPTIMIZATION OF NEURAL ARCHITECTURES IN REMOTE SENSING CLASSIFICATION PROBLEMS

AU - Coquelin, Daniel

AU - Sedona, Rocco

AU - Riedel, Morris

AU - Götz, Markus

N1 - Funding Information: This work was performed on the supercomputer ForHLR funded by the Ministry of Science, Research and the Arts Baden-Württemberg and by the Federal Ministry of Education and Research. This work is supported by the Helmholtz Association’s Initiative and Networking Funds under under the Helmholtz AI platform grant and the Helmholtz Analytics Framework collaboration. Publisher Copyright: © 2021 IEEE

PY - 2021

Y1 - 2021

N2 - BigEarthNet is one of the standard large remote sensing datasets. It has been shown previously that neural networks are effective tools to classify the image patches in this data. However, finding the optimum network hyperparameters and architecture to accurately classify the image patches in BigEarthNet remains a challenge. Searching for more accurate models manually is extremely time consuming and labour intensive. Hence, a systematic approach is advisable. One possibility is automated evolutionary Neural Architecture Search (NAS). With this NAS many of the commonly used network hyperparameters, such as loss functions, are eliminated and a more accurate network is determined.

AB - BigEarthNet is one of the standard large remote sensing datasets. It has been shown previously that neural networks are effective tools to classify the image patches in this data. However, finding the optimum network hyperparameters and architecture to accurately classify the image patches in BigEarthNet remains a challenge. Searching for more accurate models manually is extremely time consuming and labour intensive. Hence, a systematic approach is advisable. One possibility is automated evolutionary Neural Architecture Search (NAS). With this NAS many of the commonly used network hyperparameters, such as loss functions, are eliminated and a more accurate network is determined.

KW - Classification

KW - Evolutionary Algorithms

KW - NAS

KW - Neural Architecture Search

KW - Remote Sensing

UR - https://www.scopus.com/pages/publications/85129542543

U2 - 10.1109/IGARSS47720.2021.9554309

DO - 10.1109/IGARSS47720.2021.9554309

M3 - Paper

SP - 1587

EP - 1590

T2 - 2021 IEEE International Geoscience and Remote Sensing Symposium, IGARSS 2021

Y2 - 12 July 2021 through 16 July 2021

ER -

EVOLUTIONARY OPTIMIZATION OF NEURAL ARCHITECTURES IN REMOTE SENSING CLASSIFICATION PROBLEMS

Útdráttur

Ráðstefna

Athugasemd

Aðgangur að skjali

Fingerprint

Vitna í þetta