Scalable Variational Bayesian Kernel Selection for Sparse Gaussian Process Regression

Authors: Tong Teng, Jie Chen, Yehong Zhang, Bryan Kian Hsiang Low5997-6004

AAAI 2020 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental We empirically evaluate the performance of our VBKS algorithm on synthetic and massive real-world datasets.
Researcher Affiliation Academia Tong Teng,1 Jie Chen,2 Yehong Zhang,1 Bryan Kian Hsiang Low1 1Department of Computer Science, National University of Singapore, Republic of Singapore 2College of Computer Science and Software Engineering, Shenzhen University, P. R. China {tengtong, yehong, lowkh}@comp.nus.edu.sg, chenjie@szu.edu.cn
Pseudocode No The paper does not contain structured pseudocode or algorithm blocks.
Open Source Code No The paper does not provide concrete access to source code for the methodology described.
Open Datasets Yes We empirically evaluate the performance of our VBKS algorithm on synthetic and two massive real-world datasets: (a) Swissgrid dataset4 contains 210, 336 records... (b) indoor environmental quality (IEQ) dataset5 contains temperature ( C) ... 4https://www.swissgrid.ch 5http://db.csail.mit.edu/labdata/labdata.html
Dataset Splits No The paper mentions test sets but does not specify train/validation/test dataset splits needed to reproduce the experiment.
Hardware Specification Yes The real-world experiments are performed on a Linux system with 5 Nvidia Ge Force GTX 1080 GPUs.
Software Dependencies Yes Stochastic optimization for VBKS is performed in a distributed manner over the 5 GPUs using GPflow (Matthews et al. 2017).
Experiment Setup Yes In all the synthetic experiments, we use |U| = 16 inducing inputs and a batch size | D| = 32 to perform the SGA update per iteration. ... For Swissgrid Dataset. We use |U| = 800 inducing inputs and a batch size | D| = 128 for SGA update per iteration. ... For IEQ Dataset. ... We use |U| = 1000 inducing inputs and a batch size | D| = 512 for SGA update per iteration.