Understanding Probabilistic Sparse Gaussian Process Approximations
Authors: Matthias Bauer, Mark van der Wilk, Carl Edward Rasmussen
NeurIPS 2016 | Conference PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | As our main test case we use the one dimensional dataset2 considered in [7, 9] with 200 input-output pairs. Of course, sparse methods are not necessary for this toy problem, but all of the issues we raise are illustrated nicely in this one dimensional task which can easily be plotted. In Sections 3.1 to 3.3 we illustrate issues relating to the objecctive functions. We show the objective function (negative log marginal likelihood), the optimised noise σn, the negative log predictive probability and standardised mean squared error as defined in [1]. |
| Researcher Affiliation | Academia | Matthias Bauer Mark van der Wilk Carl Edward Rasmussen Department of Engineering, University of Cambridge, Cambridge, UK Max Planck Institute for Intelligent Systems, T ubingen, Germany |
| Pseudocode | No | The paper describes concepts and equations but does not contain any pseudocode or algorithm blocks. |
| Open Source Code | No | The paper does not contain any statements about releasing source code for the methodology or links to code repositories. |
| Open Datasets | Yes | As our main test case we use the one dimensional dataset2 considered in [7, 9] with 200 input-output pairs. 2Obtained from http://www.gatsby.ucl.ac.uk/~snelson/ As this behaviour is not observable in our 1D dataset, we illustrate it on the pumadyn32nm dataset4 (32 dimensions, 7168 training, 1024 test)... 4obtained from http://www.cs.toronto.edu/~delve/data/datasets.html |
| Dataset Splits | No | We consider a 4d toy dataset: 1024 training and 1024 test samples drawn from a 4d Gaussian Process... and As this behaviour is not observable in our 1D dataset, we illustrate it on the pumadyn32nm dataset4 (32 dimensions, 7168 training, 1024 test). It does not mention validation splits. |
| Hardware Specification | No | The paper does not provide specific details about the hardware used for running experiments, such as CPU/GPU models or memory specifications. |
| Software Dependencies | No | The paper does not mention any specific software dependencies with version numbers (e.g., Python version, library versions). |
| Experiment Setup | Yes | The hyperparameter θ contains the signal variance s2 f, the lengthscale ℓand the noise variance σ2 n, and is suppressed in the notation. We consider a 4d toy dataset: ... with isotropic squared exponential covariance function (l = 1.5, sf = 1) and true noise variance σ2 n = 0.01. Using a squared exponential ARD kernel with separate lengthscales for every dimension |