On the Identifiability and Interpretability of Gaussian Process Models
Authors: Jiawen Chen, Wancen Mu, Yun Li, Didong Li
NeurIPS 2023 | Conference PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | Our findings are supported by extensive simulations and real applications for both single- and multi-output settings. This work provides insight into kernel selection and interpretation for GP models, emphasizing the importance of choosing appropriate kernel structures for different tasks. Our theoretical assertions are supported by both simulations and real-world applications. |
| Researcher Affiliation | Academia | Jiawen Chen Department of Biostatistics University of North Carolina at Chapel Hill jiawenn@email.unc.edu. Wancen Mu Department of Biostatistics University of North Carolina at Chapel Hill wancen@live.unc.edu. Yun Li Department of Biostatistics University of North Carolina at Chapel Hill yun_li@med.unc.edu. Didong Li Department of Biostatistics University of North Carolina at Chapel Hill didongli@unc.edu |
| Pseudocode | No | The paper does not contain any structured pseudocode or algorithm blocks. |
| Open Source Code | No | The paper does not provide any statements about open-source code availability or links to code repositories for the described methodology. |
| Open Datasets | Yes | We employ a handwritten zero image from the MNIST dataset (Le Cun et al. (1998)) for this analysis; we extract a 20 20 section from the center of the original 100 100 image to serve as our test image. This results in a training dataset of 9600 pixels and a testing dataset with 400 pixels. In this section, we extend our comparative analysis to the widely recognized Moana Loa CO2 dataset from the Global Monitoring Laboratory s Repository as our benchmark for regression analysis (Tans and Keeling (2023)). |
| Dataset Splits | No | The paper describes training and testing splits ('varying the training sample size from 5% to 95% and keeping the test sample size as the remaining data') but does not explicitly mention a separate validation set. |
| Hardware Specification | No | The paper does not provide specific hardware details (e.g., CPU, GPU models, or memory) used for running its experiments. |
| Software Dependencies | No | The paper mentions 'optimizers such as Adam (Kingma and Ba, 2015) or stochastic gradient descent (SGD) are used to maximize the log-likelihood' but does not specify version numbers for these or any other software libraries or frameworks. |
| Experiment Setup | No | The paper describes the setup of the kernels used in simulations (e.g., 'K = P3 l=1 wl Kl, where Kl is the Matérn kernel with smoothness parameter νl being 2l 1 2 and PL l=1 wl = 1, wl 0') and sample sizes, but it does not provide specific hyperparameters like learning rates, batch sizes, or optimizer settings typically associated with model training setup. |