Agreement-Based Dynamic Active Learning with Least and Medium Certainty Query Strategy. Zhang, Y., Coutinho, E., Zhang, Z., Quan, C., & Schuller, B. In Krishnamurthy, A., Ramdas, A., Balcan, N., & Singh, A., editors, Proceedings of the 32nd International Conference on Machine Learning (ICML 2015), pages 1-5, 2015. International Machine Learning Society (IMLS). abstract bibtex In this contribution, we propose a novel method for active learning termed ‘dynamic active learn- ing’ or DAL for short, with the aim of ultimately reducing the costly human labelling work for sub- jective tasks such as speech emotion recognition. Through an adaptive query strategy, the amount of manual labelling work is minimised by deciding for each instance not only whether or not it should be annotated, but also dynamically on how many human annotators’ opinions are needed. Through extensive experiments on standardised test-beds, we show that DAL achieves the same classifica- tion accuracy of ‘traditional’ AL with a cost re- duction of up to 79.17%. Thus, the DAL method significantly improves the efficiency of existing al- gorithms, setting a new benchmark for the utmost exploitation of unlabelled data.
@inproceedings{
title = {Agreement-Based Dynamic Active Learning with Least and Medium Certainty Query Strategy},
type = {inproceedings},
year = {2015},
keywords = {article,conference},
pages = {1-5},
publisher = {International Machine Learning Society (IMLS)},
city = {Lille, France},
id = {ca15255b-57b8-3b51-9a22-035b3d36b816},
created = {2020-05-29T11:51:38.863Z},
file_attached = {true},
profile_id = {ffa9027c-806a-3827-93a1-02c42eb146a1},
last_modified = {2020-05-30T17:16:56.561Z},
read = {false},
starred = {false},
authored = {true},
confirmed = {true},
hidden = {false},
citation_key = {zhang2015agreementbasedstrategy},
source_type = {inproceedings},
folder_uuids = {aac08d0d-38e7-4f4e-a381-5271c5c099ce},
private_publication = {false},
abstract = {In this contribution, we propose a novel method for active learning termed ‘dynamic active learn- ing’ or DAL for short, with the aim of ultimately reducing the costly human labelling work for sub- jective tasks such as speech emotion recognition. Through an adaptive query strategy, the amount of manual labelling work is minimised by deciding for each instance not only whether or not it should be annotated, but also dynamically on how many human annotators’ opinions are needed. Through extensive experiments on standardised test-beds, we show that DAL achieves the same classifica- tion accuracy of ‘traditional’ AL with a cost re- duction of up to 79.17%. Thus, the DAL method significantly improves the efficiency of existing al- gorithms, setting a new benchmark for the utmost exploitation of unlabelled data.},
bibtype = {inproceedings},
author = {Zhang, Yue and Coutinho, Eduardo and Zhang, Zixing and Quan, Caijiao and Schuller, Björn},
editor = {Krishnamurthy, A and Ramdas, A and Balcan, N and Singh, A},
booktitle = {Proceedings of the 32nd International Conference on Machine Learning (ICML 2015)}
}
Downloads: 0
{"_id":"CCtPSZKNkJz2b2CWC","bibbaseid":"zhang-coutinho-zhang-quan-schuller-agreementbaseddynamicactivelearningwithleastandmediumcertaintyquerystrategy-2015","downloads":0,"creationDate":"2017-04-06T12:03:59.114Z","title":"Agreement-Based Dynamic Active Learning with Least and Medium Certainty Query Strategy","author_short":["Zhang, Y.","Coutinho, E.","Zhang, Z.","Quan, C.","Schuller, B."],"year":2015,"bibtype":"inproceedings","biburl":"https://bibbase.org/service/mendeley/ffa9027c-806a-3827-93a1-02c42eb146a1","bibdata":{"title":"Agreement-Based Dynamic Active Learning with Least and Medium Certainty Query Strategy","type":"inproceedings","year":"2015","keywords":"article,conference","pages":"1-5","publisher":"International Machine Learning Society (IMLS)","city":"Lille, France","id":"ca15255b-57b8-3b51-9a22-035b3d36b816","created":"2020-05-29T11:51:38.863Z","file_attached":"true","profile_id":"ffa9027c-806a-3827-93a1-02c42eb146a1","last_modified":"2020-05-30T17:16:56.561Z","read":false,"starred":false,"authored":"true","confirmed":"true","hidden":false,"citation_key":"zhang2015agreementbasedstrategy","source_type":"inproceedings","folder_uuids":"aac08d0d-38e7-4f4e-a381-5271c5c099ce","private_publication":false,"abstract":"In this contribution, we propose a novel method for active learning termed ‘dynamic active learn- ing’ or DAL for short, with the aim of ultimately reducing the costly human labelling work for sub- jective tasks such as speech emotion recognition. Through an adaptive query strategy, the amount of manual labelling work is minimised by deciding for each instance not only whether or not it should be annotated, but also dynamically on how many human annotators’ opinions are needed. Through extensive experiments on standardised test-beds, we show that DAL achieves the same classifica- tion accuracy of ‘traditional’ AL with a cost re- duction of up to 79.17%. Thus, the DAL method significantly improves the efficiency of existing al- gorithms, setting a new benchmark for the utmost exploitation of unlabelled data.","bibtype":"inproceedings","author":"Zhang, Yue and Coutinho, Eduardo and Zhang, Zixing and Quan, Caijiao and Schuller, Björn","editor":"Krishnamurthy, A and Ramdas, A and Balcan, N and Singh, A","booktitle":"Proceedings of the 32nd International Conference on Machine Learning (ICML 2015)","bibtex":"@inproceedings{\n title = {Agreement-Based Dynamic Active Learning with Least and Medium Certainty Query Strategy},\n type = {inproceedings},\n year = {2015},\n keywords = {article,conference},\n pages = {1-5},\n publisher = {International Machine Learning Society (IMLS)},\n city = {Lille, France},\n id = {ca15255b-57b8-3b51-9a22-035b3d36b816},\n created = {2020-05-29T11:51:38.863Z},\n file_attached = {true},\n profile_id = {ffa9027c-806a-3827-93a1-02c42eb146a1},\n last_modified = {2020-05-30T17:16:56.561Z},\n read = {false},\n starred = {false},\n authored = {true},\n confirmed = {true},\n hidden = {false},\n citation_key = {zhang2015agreementbasedstrategy},\n source_type = {inproceedings},\n folder_uuids = {aac08d0d-38e7-4f4e-a381-5271c5c099ce},\n private_publication = {false},\n abstract = {In this contribution, we propose a novel method for active learning termed ‘dynamic active learn- ing’ or DAL for short, with the aim of ultimately reducing the costly human labelling work for sub- jective tasks such as speech emotion recognition. Through an adaptive query strategy, the amount of manual labelling work is minimised by deciding for each instance not only whether or not it should be annotated, but also dynamically on how many human annotators’ opinions are needed. Through extensive experiments on standardised test-beds, we show that DAL achieves the same classifica- tion accuracy of ‘traditional’ AL with a cost re- duction of up to 79.17%. Thus, the DAL method significantly improves the efficiency of existing al- gorithms, setting a new benchmark for the utmost exploitation of unlabelled data.},\n bibtype = {inproceedings},\n author = {Zhang, Yue and Coutinho, Eduardo and Zhang, Zixing and Quan, Caijiao and Schuller, Björn},\n editor = {Krishnamurthy, A and Ramdas, A and Balcan, N and Singh, A},\n booktitle = {Proceedings of the 32nd International Conference on Machine Learning (ICML 2015)}\n}","author_short":["Zhang, Y.","Coutinho, E.","Zhang, Z.","Quan, C.","Schuller, B."],"editor_short":["Krishnamurthy, A.","Ramdas, A.","Balcan, N.","Singh, A."],"biburl":"https://bibbase.org/service/mendeley/ffa9027c-806a-3827-93a1-02c42eb146a1","bibbaseid":"zhang-coutinho-zhang-quan-schuller-agreementbaseddynamicactivelearningwithleastandmediumcertaintyquerystrategy-2015","role":"author","urls":{},"keyword":["article","conference"],"metadata":{"authorlinks":{"coutinho, e":"https://bibbase.org/service/mendeley/ffa9027c-806a-3827-93a1-02c42eb146a1"}},"downloads":0},"search_terms":["agreement","based","dynamic","active","learning","medium","certainty","query","strategy","zhang","coutinho","zhang","quan","schuller"],"keywords":["article","conference"],"authorIDs":["mo4CFXJ7ukAMT9nho"],"dataSources":["Tcd3cXtdQsiKHPZsW","ya2CyA73rpZseyrZ8","YqW8pMoihb7JazZcx","HEzruaK9ZuLGaJrsP","jDf6YRbdcAMnNSFh6","2252seNhipfTmjEBQ"]}