Title | ||
---|---|---|
Sequential Randomized Matrix Factorization for Gaussian Processes: Efficient Predictions and Hyper-parameter Optimization. |
Abstract | ||
---|---|---|
This paper presents a sequential randomized lowrank matrix factorization approach for incrementally predicting values of an unknown function at test points using the Gaussian Processes framework. It is well-known that in the Gaussian processes framework, the computational bottlenecks are the inversion of the (regularized) kernel matrix and the computation of the hyper-parameters defining the kernel. The main contributions of this paper are two-fold. First, we formalize an approach to compute the inverse of the kernel matrix using randomized matrix factorization algorithms in a streaming scenario, i.e., data is generated incrementally over time. The metrics of accuracy and computational efficiency of the proposed method are compared against a batch approach based on use of randomized matrix factorization and an existing streaming approach based on approximating the Gaussian process by a finite set of basis vectors. Second, we extend the sequential factorization approach to a class of kernel functions for which the hyperparameters can be efficiently optimized. All results are demonstrated on two publicly available datasets. |
Year | Venue | Field |
---|---|---|
2017 | arXiv: Learning | Kernel (linear algebra),Mathematical optimization,Finite set,Hyperparameter,Matrix decomposition,Gaussian process,Factorization,Artificial intelligence,Basis (linear algebra),Machine learning,Mathematics,Kernel (statistics) |
DocType | Volume | Citations |
Journal | abs/1711.06989 | 0 |
PageRank | References | Authors |
0.34 | 3 | 2 |
Name | Order | Citations | PageRank |
---|---|---|---|
Shaunak D. Bopardikar | 1 | 1 | 5.48 |
George S. Eskander Ekladious | 2 | 0 | 0.68 |