TY - GEN
T1 - Bayesian Optimization in High-Dimensional Spaces
T2 - 12th International Conference on Information, Intelligence, Systems and Applications, IISA 2021
AU - Malu, Mohit
AU - Dasarathy, Gautam
AU - Spanias, Andreas
N1 - Funding Information:
This project was supported in part by the NSF Awards 1540040, 2003111, 2048223 and the SenSIP Center.
Publisher Copyright:
© 2021 IEEE.
PY - 2021/7/12
Y1 - 2021/7/12
N2 - Bayesian optimization (BO) has been widely applied to several modern science and engineering applications such as machine learning, neural networks, robotics, aerospace engineering, experimental design. BO has emerged as the modus operandi for global optimization of an arbitrary expensive to evaluate black box function f. Although BO has been very successful in low dimensions, scaling it to high dimensional spaces has been significantly challenging due to its exponentially increasing statistical and computational complexity with increasing dimensions. In this era of high dimensional data where the input features are of million dimensions scaling BO to higher dimensions is one of the important goals in the field. There has been a lot of work in recent years to scale BO to higher dimensions, in many of these methods some underlying structure on the objective function is exploited. In this paper, we review recent efforts in this area. In particular, we focus on the methods that exploit different underlying structures on the objective function to scale BO to high dimensions.
AB - Bayesian optimization (BO) has been widely applied to several modern science and engineering applications such as machine learning, neural networks, robotics, aerospace engineering, experimental design. BO has emerged as the modus operandi for global optimization of an arbitrary expensive to evaluate black box function f. Although BO has been very successful in low dimensions, scaling it to high dimensional spaces has been significantly challenging due to its exponentially increasing statistical and computational complexity with increasing dimensions. In this era of high dimensional data where the input features are of million dimensions scaling BO to higher dimensions is one of the important goals in the field. There has been a lot of work in recent years to scale BO to higher dimensions, in many of these methods some underlying structure on the objective function is exploited. In this paper, we review recent efforts in this area. In particular, we focus on the methods that exploit different underlying structures on the objective function to scale BO to high dimensions.
UR - http://www.scopus.com/inward/record.url?scp=85117437236&partnerID=8YFLogxK
UR - http://www.scopus.com/inward/citedby.url?scp=85117437236&partnerID=8YFLogxK
U2 - 10.1109/IISA52424.2021.9555522
DO - 10.1109/IISA52424.2021.9555522
M3 - Conference contribution
AN - SCOPUS:85117437236
T3 - IISA 2021 - 12th International Conference on Information, Intelligence, Systems and Applications
BT - IISA 2021 - 12th International Conference on Information, Intelligence, Systems and Applications
PB - Institute of Electrical and Electronics Engineers Inc.
Y2 - 12 July 2021 through 14 July 2021
ER -