Efficient methods for Gaussian Markov random fields under sparse linear constraints
Authors: David Bolin, Jonas Wallin
NeurIPS 2021 | Conference PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | The methods are illustrated numerically in Section 6. A discussion closes the article, which is supported by a supplementary materials containing proofs and technical details. 6 Numerical illustrations |
| Researcher Affiliation | Academia | David Bolin King Abdullah University of Science and Technology david.bolin@kaust.edu.sa Jonas Wallin Department of Statistics, Lund University jonas.wallin@stat.lu.se |
| Pseudocode | Yes | Algorithm 1 Constraint basis construction. Require: A (a k n matrix of rank k) ... Algorithm 2 CB(A). Constraint basis construction for non-overlapping subsets of constraints. ... Algorithm 3 Sampling X N µ, Q 1 subject to AX = b. ... Algorithm 4 Sampling X N µ, Q 1 subject to AX = b and Y = y where Y N BX, σ2 Y I . |
| Open Source Code | Yes | timings are obtained through R [25] implementations, available in the CB R package [6] |
| Open Datasets | No | The paper uses "simulated data" and does not provide concrete access information (link, DOI, repository, or formal citation) for a publicly available or open dataset. |
| Dataset Splits | No | The paper does not provide specific dataset split information (exact percentages, sample counts, citations to predefined splits, or detailed splitting methodology) needed to reproduce the data partitioning. It mentions using "simulated data" and "randomly selected locations" without specifying splits. |
| Hardware Specification | Yes | run on an i Mac Pro computer with a 3.2 GHz Intel Xeon processor. |
| Software Dependencies | Yes | timings are obtained through R [25] implementations, available in the CB R package [6] |
| Experiment Setup | Yes | The observations are simulated using the parameters κ2 = 0.5, φ = 1 and α = 2 or α = 4. ... Following [13], we choose a = 0.01 and σ2 = 10 4 and generate 50 observations at randomly selected locations in [0, 4] [0, 4] and predict the function f at N 2 = 202 regularly spaced locations in the square. Independent Matérn priors with α = 4 are assumed for f1 and f2 ... a regular triangulation of an extended domain [ 2, 6] [ 2, 6] ... only enforce the divergence constraint at every third node for the SPDE model. |