Organization
Faculty of Business Data Science Associate Professor
Title
Associate Professor
External link
Holland, Matthew
Degree
-
Ph.D ( Nara Institute of Science and Technology )
Research Interests
-
Statistical learning theory
-
Machine learning
Papers
-
Soft ascent-descent as a stable and flexible alternative to flooding Reviewed
Matthew J. Holland, Kosuke Nakatani
Advances in Neural Information Processing Systems 37 (NeurIPS 2024) 2024.12
-
Making Robust Generalizers Less Rigid with Soft Ascent-Descent
Matthew J. Holland, Toma Hamada
arXiv preprint 2024.8
-
Criterion Collapse and Loss Distribution Control Reviewed
Matthew J. Holland
Proceedings of Machine Learning Research (ICML 2024) 235 18547 - 18567 2024.7
-
Robust variance-regularized risk minimization with concomitant scaling Reviewed
Matthew J. Holland
Proceedings of Machine Learning Research (AISTATS 2024) 238 1144 - 1152 2024.5
-
A Survey of Learning Criteria Going Beyond the Usual Risk Reviewed
Matthew J. Holland, Kazuki Tanabe
Journal of Artificial Intelligence Research 78 781 - 821 2023.11
-
Flexible risk design using bi-directional dispersion Reviewed
Matthew J. Holland
Proceedings of Machine Learning Research (AISTATS 2023) 206 1586 - 1623 2023.4
-
Learning with risks based on M-location Reviewed
Matthew J. Holland
Machine Learning 111 4679 - 4718 2022.8
-
Spectral risk-based learning using unbounded losses Reviewed
Matthew J. Holland, El Mehdi Haress
Proceedings of Machine Learning Research (AISTATS 2022) 151 ( 1 ) 1871 - 1886 2022.3
-
Anytime Guarantees under Heavy-Tailed Data Reviewed
Matthew J. Holland
Proceedings of the AAAI Conference on Artificial Intelligence (AAAI 2022) 36 ( 6 ) 6918 - 6925 2022.2
-
Making learning more transparent using conformalized performance prediction Reviewed
Matthew J. Holland
ICML 2021, Workshop on Distribution-Free Uncertainty Quantification 2021.7
-
Robustness and scalability under heavy tails, without strong convexity Reviewed
Matthew J. Holland
Proceedings of Machine Learning Research (AISTATS 2021) 130 ( 1 ) 2021.4
-
Learning with risk-averse feedback under potentially heavy tails Reviewed
Matthew J. Holland, El Mehdi Haress
Proceedings of Machine Learning Research (AISTATS 2021) 130 ( 1 ) 2021.4
-
Scaling-Up Robust Gradient Descent Techniques Reviewed
Matthew J. Holland
Proceedings of the AAAI Conference on Artificial Intelligence (AAAI 2021) 35 ( 9 ) 7694 - 7701 2021.2
-
Better scalability under potentially heavy-tailed feedback
Matthew J. Holland
Archival version 2020.12
-
PAC-Bayes under potentially heavy tails Reviewed
Matthew J. Holland
Advances in Neural Information Processing Systems 32 (NeurIPS 2019) 2019.12
-
Efficient learning with robust gradient descent Reviewed
Matthew J. Holland, Kazushi Ikeda
Machine Learning 108 ( 8 ) 1523 - 1560 2019.9
-
Better generalization with less data using robust gradient descent Reviewed
Matthew J. Holland, Kazushi Ikeda
Proceedings of Machine Learning Research (ICML 2019) 97 2761 - 2770 2019.6
-
Distribution-robust mean estimation via smoothed random perturbations
Matthew J. Holland
arXiv preprint 2019.6
-
Robust descent using smoothed multiplicative noise Reviewed
Matthew J. Holland
Proceedings of Machine Learning Research (AISTATS 2019) 89 ( 1 ) 703 - 711 2019.4
-
Classification using margin pursuit Reviewed
Matthew J. Holland
Proceedings of Machine Learning Research (AISTATS 2019) 89 ( 1 ) 712 - 720 2019.4
-
Robust regression using biased objectives Reviewed
Matthew J. Holland, Kazushi Ikeda
Machine Learning 106 ( 9-10 ) 1643 - 1679 2017.10
-
Minimum Proper Loss Estimators for Parametric Models Reviewed
Matthew J. Holland, Kazushi Ikeda
IEEE Transactions on Signal Processing 64 ( 3 ) 704 - 713 2016.2
-
Location robust estimation of predictive Weibull parameters in short-term wind speed forecasting Reviewed
Matthew J. Holland, Kazushi Ikeda
Proceedings of ICASSP 2015 2015.4
-
Forecasting in wind energy applications with site-adaptive Weibull estimation Reviewed
Matthew J. Holland, Kazushi Ikeda
Proceedings of ICASSP 2014 2014.5
MISC
-
Matthew J. Holland
IPSJ Magazine 60 ( 8 ) 780 - 781 2019.7
-
Robust gradient descent via back-propagation: A Chainer-based tutorial
Matthew J. Holland
2019.5
Presentations
-
Proposed learning method assuming distribution shift due to data corruption
Toma Hamada, Matthew J. Holland
The 38th Annual Conference of the Japan Society for Artificial Intelligence (JSAI 2024) 2024.5
Works
-
addro: learning with concentrated OCE
Matthew J. Holland
2024.7
-
collapse: a study of criterion collapse in machine learning
Matthew J. Holland
2024.2
-
bdd-flood: links between bi-directional dispersion and flooding
Matthew J. Holland
2023.10
-
bdd-mv: mean-variance minimization using bi-directional dispersion functions
Matthew J. Holland
2023.1
-
offgen: a visual "explainer" for off-sample generalization metrics
Matthew J. Holland
2022.7
-
bdd: bidirectional dispersion and risk function design
Matthew J. Holland
2022.3
-
spectral: learning based on spectral risks
2021.5
-
anytime: robust learning anytime
Matthew J. Holland
2021.5
-
mrisk: learning with risks based on M-location
Matthew J. Holland
2021.4
-
sgd-roboost: robust confidence boosting of SGD sub-processes
Matthew J. Holland
2020.12
-
1dim (Python): estimators using smoothed random perturbations
Matthew J. Holland
2019.6
-
rgd (Python): Robust gradient descent examples
Matthew J. Holland
2019.5
-
catcube (Python): for new margin-based learning algorithms
Matthew J. Holland
2019.4
-
dosha-mesh (Python): Processing JMA landslide risk data
Matthew J. Holland
2018.8
Awards
-
Expert Reviewer Recognition (2024)
2024.8 Transactions on Machine Learning Research (TMLR)
Matthew J. Holland
-
Expert Reviewer Recognition (2023)
2023.7 Transactions on Machine Learning Research (TMLR)
Matthew J. Holland
-
Distinguished Presentation Award (NEURO2022)
2022.7 Japanese Neural Network Society
Matthew J. Holland
-
Young Researcher Award (JNNS2021)
2021.9 Japanese Neural Network Society
Matthew J. Holland
-
2019.11 IEICE, Information-Based Induction Sciences and Machine Learning (IBISML) Technical Committee
Matthew J. Holland
-
NAIST Top Student Award (PhD Program)
2018.3 Foundation for NAIST
Matthew James Holland
-
Student Paper Award
2017.2 IEEE Kansai Section
Matthew James Holland
-
NAIST Top Student Award (Master's Program)
2015.3 Foundation for NAIST
Matthew James Holland
Research Projects
-
Novel learning algorithms through off-sample generalization metric design
Grant number：23K24902 2022.4 - 2026.3
Japan Society for the Promotion of Science Grant-in-Aid for Scientific Research (B)
Matthew J. Holland
Authorship：Principal investigator Grant type：Competitive
-
Designing new learning algorithms via novel feedback generation mechanisms
2022.1 - 2022.12
Japanese Neural Network Society JNNS30 Commemorative Research Grant
Matthew J. Holland
Authorship：Principal investigator
-
Machine learning with guarantees under diverse risk measures
Grant number：JPMJPR21C6 2021.10 - 2026.3
JST PRESTO (Fundamental Technologies for Trustworthy AI)
Matthew J. Holland
Authorship：Principal investigator Grant type：Competitive
-
Learning with guarantees under more diverse notions of risk
Grant number：JPMJAX200O 2020.11 - 2022.3
JST ACT-X (Frontier of Mathematics and Information Science)
Matthew J. Holland
Authorship：Principal investigator Grant type：Competitive
-
Robust and efficient learning algorithms through control of margin distributions
Grant number：19K20342 2019.4 - 2022.3
Japan Society for Promotion of Science Grant-in-Aid for Early-Career Scientists
Matthew J. Holland
Authorship：Principal investigator Grant type：Competitive
-
2019.1 - 2020.12
Kayamori Foundation for Information Science Advancement Research grant
Matthew J. Holland
Authorship：Principal investigator Grant type：Competitive
-
Robust and efficient learning algorithms through control of margin distributions
Grant number：18H06477 2018.8 - 2019.3
Japan Society for the Promotion of Science Grant-in-Aid for Research Activity start-up
Matthew James Holland
Authorship：Principal investigator Grant type：Competitive
-
Safe AI is efficient AI: improved generalization via robust learning algorithms
Grant number：JPMJPR17UO 2017.10 - 2019.3
Japan Science and Technology Agency ACT-I (Information and Future)
Matthew James Holland
Authorship：Principal investigator Grant type：Competitive
-
Analysis of automated loss-selecting algorithms and their extension to time-series data
Grant number：16J06659 2016.4 - 2017.12
Japan Society for the Promotion of Science Grant-in-Aid for JSPS Fellows
Matthew James Holland
Authorship：Principal investigator Grant type：Competitive