Non-Stationary Spectral Kernels
Authors: Sami Remes, Markus Heinonen, Samuel Kaski
NeurIPS 2017 | Conference PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | We show with case studies that these kernels are necessary when modelling even rather simple time series, image or geospatial data with non-stationary characteristics.We show the expressivity of the kernel with experiments on time series data, image-based pattern recognition and extrapolation, and on climate data modelling. |
| Researcher Affiliation | Academia | sami.remes@aalto.fi Markus Heinonen markus.o.heinonen@aalto.fi Samuel Kaski samuel.kaski@aalto.fi Helsinki Institute for Information Technology HIIT Department of Computer Science, Aalto University |
| Pseudocode | No | The paper does not contain structured pseudocode or algorithm blocks. |
| Open Source Code | Yes | Implementation available at https://github.com/sremes/nonstationary-spectral-kernels |
| Open Datasets | Yes | NASA provides a land surface temperature dataset that we used to demonstrate our kernel in analysis of spatio-temporal data. |
| Dataset Splits | No | The paper describes using a "training data" subset and later making "test data predictions" for the land surface temperature dataset, and for image data, it uses "majority of the image as training data" to predict a "missing cross-section" and "extrapolate". However, it does not specify exact split percentages, sample counts for each split, or explicit mention of a separate validation set for reproducibility. |
| Hardware Specification | No | The paper mentions "computational resources provided by the Aalto Science-IT project" but does not provide specific hardware details like GPU/CPU models, processor types, or memory amounts used for running the experiments. |
| Software Dependencies | No | We employ the GPML Matlab toolbox, which directly implements the SM and SE kernels, and the SS kernel as a meta kernel combining simple cosine kernels. We implemented the proposed GSM kernel and inference in Matlab. The paper mentions software like "Matlab" and "GPML Matlab toolbox" but does not specify any version numbers for these software components. |
| Experiment Setup | Yes | For optimising the log posterior (10) we employ the L-BFGS algorithm. For both our method and the comparisons, we restart the optimisation from 10 different initialisations, each of which is chosen as the best among 100 randomly sampled hyperparameter values as evaluating the log posterior is cheap compared to evaluating gradients or running the full optimisation. For GSM, SM and SS we used Q = 5 mixture components for the metal texture, and Q = 10 components for the more complex wood texture. |