Abstract
Gaussian process (GP) is a Bayesian nonparametric regression model, showing good performance in various applications. However, its hyperparameterestimation procedure suffers from numerous covariance-matrix inversions of prohibitively O(N3) operations. In this paper, we propose using the quasi-Newton BFGS O(N2)-operation formula to update recursively the inverse of covariance matrix at every iteration. As for the involved log det computation, a power-series expansion based approximation and compensation scheme is proposed with only 50N2 operations. A number of numerical tests are performed based on the 2D- sinusoidal regression example and the Wiener-Hammerstein identification example. It is shown that by using the proposed implementation, more than 80% O(N3) operations are eliminated, and the speedup of 5 - 9 can be achieved.
Original language | English |
---|---|
Title of host publication | Proceedings of the 16th IFAC World Congress, 2005 |
Pages | 217-217 |
Number of pages | 1 |
Volume | 16 |
DOIs | |
Publication status | Published - 8 Jul 2005 |
Event | 16th IFAC World Congress Conference - Prague, Czech Republic Duration: 4 Jul 2005 → 8 Jul 2005 |
Conference
Conference | 16th IFAC World Congress Conference |
---|---|
Country | Czech Republic |
City | Prague |
Period | 4/07/05 → 8/07/05 |
Keywords
- Gaussian process regression
- compensation
- approximation
- power series expansion
- matrix inverse