A Region-Based Model for Estimating Urban Air Pollution
Authors: Arnaud Jutzeler, Jason Li, Boi Faltings
AAAI 2014 | Conference PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | In this paper, we propose a novel region-based Gaussian process model for estimating urban air pollution dispersion, and applied it to a large dataset of ultrafine particle (UFP) measurements collected from a network of sensors located on several trams in the city of Zurich. We show that compared to existing grid-based models, the regionbased model produces better predictions across aggregates of all time scales. |
| Researcher Affiliation | Academia | Arnaud Jutzeler1, Jason Jingshi Li2, Boi Faltings1 1Ecole Polytechnique Federale de Lausanne 2The Australian National University |
| Pseudocode | No | The paper describes methods and equations but does not contain a clearly labeled pseudocode or algorithm block. |
| Open Source Code | No | The paper states, 'We implemented our own java-based platform to carry out GPR. We coded it nearly from scratch using only EJML1 as linear algebra library.' However, it does not provide any link or explicit statement about the code being open-source or publicly available. |
| Open Datasets | Yes | A more complete description of the dataset can be found in (Li et al. 2012; Hasenfratz et al. 2014). |
| Dataset Splits | Yes | Models are validated through standard random 10-fold cross-validation. |
| Hardware Specification | Yes | It took roughly 60 hours on a 64-cores AMD Opteron 6272 @2.1Ghz to run all the evaluation tests of next sections which gathered more than 14,000 fitting tasks with a maximum of 500 iterations each on 180-points training sets. |
| Software Dependencies | Yes | We coded it nearly from scratch using only EJML1 as linear algebra library. The conjugate gradient optimizer was taken from the Matlab toolbox GPML v.2 (Rasmussen and Nickisch 2010) and translated in Java. |
| Experiment Setup | Yes | Our GP prior models contain several hyper-parameters θ = (c, σ2 f LU , σ2 f S, ℓS, ℓLU, σ2 n) whose values are not known a priori. Those hyperparameters were learned during the regression using the same training data by maximizing the marginal likelihood (ML)... A non-linear conjugate gradient optimizer was used to carry out that task. |