WEKO3
アイテム
{"_buckets": {"deposit": "31835932-f935-447e-be19-48d5c5e1b732"}, "_deposit": {"id": "5359", "owners": [1], "pid": {"revision_id": 0, "type": "depid", "value": "5359"}, "status": "published"}, "_oai": {"id": "oai:nied-repo.bosai.go.jp:00005359", "sets": []}, "author_link": [], "item_10001_biblio_info_7": {"attribute_name": "書誌情報", "attribute_value_mlt": [{"bibliographicIssueDates": {"bibliographicIssueDate": "2021-11-01"}, "bibliographicIssueNumber": "6", "bibliographicPageEnd": "3564", "bibliographicPageStart": "3552", "bibliographicVolumeNumber": "92", "bibliographic_titles": [{"bibliographic_title": "Seismological Research Letters", "bibliographic_titleLang": "en"}]}]}, "item_10001_description_5": {"attribute_name": "抄録", "attribute_value_mlt": [{"subitem_description": "Abstract\n\n Data-driven machine-learning approaches are being increasingly applied to construct empirical ground-motion models (GMMs). It is a standard practice to divide observational records into learning and test datasets to correctly evaluate the predictive performance of a developed model. However, in this study, we show that division based on records or earthquakes is inappropriate for evaluating the generalization performance on recorded sites when GMMs include site-condition proxies as input variables. Complex models exhibit small residuals at sites used in the training process, but exhibit large residuals at new sites owing to overfitting to the trained sites. As a simple solution, we propose a neural network model that has monotonic dependence on some of the input variables. The model successfully obtains the generalization performance on recorded sites, although it lacks ability to represent oversaturation with input variables suggested in extreme ground-motion ranges. Therefore, alternative methods should be investigated to develop robust data-driven models under general conditions. Dividing the sites into learning and test data would play a fundamental role in developing such robust models.", "subitem_description_language": "en", "subitem_description_type": "Other"}]}, "item_10001_publisher_8": {"attribute_name": "出版者", "attribute_value_mlt": [{"subitem_publisher": "Seismological Society of America (SSA)", "subitem_publisher_language": "en"}]}, "item_10001_relation_14": {"attribute_name": "DOI", "attribute_value_mlt": [{"subitem_relation_type_id": {"subitem_relation_type_id_text": "10.1785/0220210099"}}]}, "item_10001_source_id_9": {"attribute_name": "ISSN", "attribute_value_mlt": [{"subitem_source_identifier": "1938-2057", "subitem_source_identifier_type": "EISSN"}]}, "item_creator": {"attribute_name": "著者", "attribute_type": "creator", "attribute_value_mlt": [{"creatorNames": [{"creatorName": "Tomohisa Okazaki", "creatorNameLang": "en"}]}, {"creatorNames": [{"creatorName": "Nobuyuki Morikawa", "creatorNameLang": "en"}]}, {"creatorNames": [{"creatorName": "Hiroyuki Fujiwara", "creatorNameLang": "en"}]}, {"creatorNames": [{"creatorName": "Naonori Ueda", "creatorNameLang": "en"}]}]}, "item_language": {"attribute_name": "言語", "attribute_value_mlt": [{"subitem_language": "eng"}]}, "item_title": "Monotonic Neural Network for Ground-Motion Predictions to Avoid Overfitting to Recorded Sites", "item_titles": {"attribute_name": "タイトル", "attribute_value_mlt": [{"subitem_title": "Monotonic Neural Network for Ground-Motion Predictions to Avoid Overfitting to Recorded Sites", "subitem_title_language": "en"}]}, "item_type_id": "40001", "owner": "1", "path": ["1670839190650"], "permalink_uri": "https://nied-repo.bosai.go.jp/records/5359", "pubdate": {"attribute_name": "PubDate", "attribute_value": "2023-03-30"}, "publish_date": "2023-03-30", "publish_status": "0", "recid": "5359", "relation": {}, "relation_version_is_last": true, "title": ["Monotonic Neural Network for Ground-Motion Predictions to Avoid Overfitting to Recorded Sites"], "weko_shared_id": -1}
Monotonic Neural Network for Ground-Motion Predictions to Avoid Overfitting to Recorded Sites
https://nied-repo.bosai.go.jp/records/5359
https://nied-repo.bosai.go.jp/records/5359d8e33117-2278-4e4e-895b-495d43f54681
Item type | researchmap(1) | |||||||||||||
---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
公開日 | 2023-03-30 | |||||||||||||
タイトル | ||||||||||||||
言語 | en | |||||||||||||
タイトル | Monotonic Neural Network for Ground-Motion Predictions to Avoid Overfitting to Recorded Sites | |||||||||||||
言語 | ||||||||||||||
言語 | eng | |||||||||||||
著者 |
Tomohisa Okazaki
× Tomohisa Okazaki
× Nobuyuki Morikawa
× Hiroyuki Fujiwara
× Naonori Ueda
|
|||||||||||||
抄録 | ||||||||||||||
内容記述タイプ | Other | |||||||||||||
内容記述 | Abstract Data-driven machine-learning approaches are being increasingly applied to construct empirical ground-motion models (GMMs). It is a standard practice to divide observational records into learning and test datasets to correctly evaluate the predictive performance of a developed model. However, in this study, we show that division based on records or earthquakes is inappropriate for evaluating the generalization performance on recorded sites when GMMs include site-condition proxies as input variables. Complex models exhibit small residuals at sites used in the training process, but exhibit large residuals at new sites owing to overfitting to the trained sites. As a simple solution, we propose a neural network model that has monotonic dependence on some of the input variables. The model successfully obtains the generalization performance on recorded sites, although it lacks ability to represent oversaturation with input variables suggested in extreme ground-motion ranges. Therefore, alternative methods should be investigated to develop robust data-driven models under general conditions. Dividing the sites into learning and test data would play a fundamental role in developing such robust models. |
|||||||||||||
言語 | en | |||||||||||||
書誌情報 |
en : Seismological Research Letters 巻 92, 号 6, p. 3552-3564 |
|||||||||||||
出版者 | ||||||||||||||
言語 | en | |||||||||||||
出版者 | Seismological Society of America (SSA) | |||||||||||||
ISSN | ||||||||||||||
収録物識別子タイプ | EISSN | |||||||||||||
収録物識別子 | 1938-2057 | |||||||||||||
DOI | ||||||||||||||
関連識別子 | 10.1785/0220210099 |