Variational Inference for Nonparametric Bayesian Quantile Regression
Authors: Sachinthaka Abeywardana, Fabio Ramos
AAAI 2015 | Conference PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | Our algorithm is also tested in four real world examples. In the motorcycle dataset, acceleration experienced by a helmet in a crash is measured over time with the goal of interpolating between existing measurements. This is a popular dataset to assess heteroscedastic inference methods. In the bone density dataset, the goal is to predict the bone density of individuals as a function of age. The birth weight dataset aims to predict infants weight as a function of the mothers age and weight. Finally, the snow fall dataset, attempts to predict snow fall at Fort Collins in January, as a function of snow fall in September-December. We have used 80% of the data as training and the rest as testing and iterated over 20 times for each experiment. The cases were randomly permuted in each iteration. The proposed method is compared against its nearest competitor, the EP approach, Heteroscedastic Quantile Gaussian Processes (HQGP) as well as, against a linear method (Lin) which attempts to find the quantile as a polynomial function of the inputs (polynomial basis function, in this case having fα = β0 + β1x + β1x2 + ... + β7x7). |
| Researcher Affiliation | Academia | Sachinthaka Abeywardana School of Information Technologies University of Sydney NSW 2006, Australia sachinra@it.usyd.edu.au Fabio Ramos School of Information Technologies University of Sydney NSW 2006, Australia fabio.ramos@sydney.edu.au |
| Pseudocode | No | The paper does not contain any sections or figures explicitly labeled 'Pseudocode' or 'Algorithm'. |
| Open Source Code | Yes | Code and data are available at http://www.bitbucket.org/sachinruk/gpquantile |
| Open Datasets | Yes | Code and data are available at http://www.bitbucket.org/sachinruk/gpquantile |
| Dataset Splits | No | The paper states, 'We have used 80% of the data as training and the rest as testing and iterated over 20 times for each experiment.' but does not explicitly mention a separate validation split. |
| Hardware Specification | No | The paper does not specify any particular hardware details such as CPU/GPU models or cloud instance types used for running the experiments. |
| Software Dependencies | No | The paper mentions software components and methods like Python, GP, EP, HQGP, and Lin, but does not provide specific version numbers for any libraries or dependencies. |
| Experiment Setup | Yes | The square exponential kernel was used in evaluating the VB, EP and HQGP methods. In the case of the real world datasets, the output is standardised to have zero mean and unit variance so that comparisons could be made across datasets. |