Efficient large scale kernel ridge regression via ensemble SPSD approximation

Tang, Xiaoliang; Wang, Xing; Di, Jia

doi:10.1145/3194206.3194236

Cited by 1 publication

(2 citation statements)

References 9 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Recently, researchers have developed kinds of online learning algorithms to track the dynamic changes [5]- [9]. As nonlinear regression problem is concerned, different kinds of kernel based online learning algorithms have been developed [13], [14], [30]- [33]. Thereinto, Haworth et al [30] introduce local temporal windows to separate the whole kernel matrix of KRR into several smaller local temporal kernel matrices, and incorporate the new sample to one local temporal kernel incrementally which is computationally more efficient than a single large kernel.…”

Section: Related Workmentioning

confidence: 99%

See 1 more Smart Citation

Toward Budgeted Online Kernel Ridge Regression on Streaming Data

et al. 2019

View full text Add to dashboard Cite

''Concept drift'' makes learning from streaming data fundamentally different from traditional batch learning. Focusing on the regression task on streaming data, this paper presents an efficient online learning algorithm, i.e., budgeted online kernel ridge regression (BOKRR). It is a budget version kernel ridge regression algorithm coupled with minimum contribution criterion to maintain the budget of the active set. BOKRR employs low-rank correction technology and the Sherman-Morrison-Woodbury formula to update the dynamic KRR model with the computational complexity of only O(B 2) with B learning samples (Budget size of an active set). Limited storage burden and efficient computational ability make the proposed BOKRR algorithm an ideal candidate to process streaming data. The experimental results on benchmark and real-world datasets further demonstrate the validity and efficiency of the proposed algorithms. INDEX TERMS Budget, kernel ridge regression, online learning, streaming data.

show abstract

Section: Related Workmentioning

confidence: 99%

“…Kernel least mean square (KLMS) [14] is the simplest kernel adaptive filter, which is easy to implement and effective for learning complex systems. Tang et al [33] propose an efficient large scale KRR based on ensemble symmetric positive semi-definite (SPSD) approximation method.…”

Section: Related Workmentioning

confidence: 99%