Gaussian Process-Based Real-Time Learning for Safety Critical Applications
Authors: Armin Lederer, Alejandro J Ordóñez Conejo, Korbinian A Maier, Wenxin Xiao, Jonas Umlauft, Sandra Hirche
ICML 2021 | Conference PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | The efficiency and performance of the proposed real-time learning approach is demonstrated in a comparison to state-of-the-art methods. ... We evaluate the performance of the Lo G-GP approach on three real-world data sets. The SARCOS data set ... buzz in social media data set ... and the individual household electric power consumption data set... |
| Researcher Affiliation | Academia | 1Department of Electrical and Computer Engineering, Technical University of Munich, Munich, Germany 2Tecnológico de Costa Rica, Cartago, Costa Rica 3Department of Computer Science and Technology, Peking University, Beijing, China. |
| Pseudocode | Yes | Algorithm 1 Updating of a Lo G-GP; Algorithm 2 Predicting with a Lo G-GP |
| Open Source Code | Yes | Matlab code is online available at https://gitlab.lrz. de/alederer/Log-GP. |
| Open Datasets | Yes | The SARCOS data set (Rasmussen & Williams, 2006) contains 44484 samples... Moreover, we employ the buzz in social media data set (Douzal-chouakria et al., 2013), which consists of 583250 samples... and the individual household electric power consumption data set (Dua & Graff, 2017) composed of 2048380 measurements... |
| Dataset Splits | No | The paper describes using "the first 1000 training samples for all methods" for hyperparameter fitting and then evaluating in a "sequential setting", but it does not specify explicit train/validation/test dataset splits with percentages or fixed sample counts required for reproduction. |
| Hardware Specification | No | The paper does not provide specific details about the hardware used to run the experiments. |
| Software Dependencies | No | The paper mentions "Matlab code" but does not provide specific version numbers for Matlab or any other software dependencies. |
| Experiment Setup | Yes | The Lo G-GP is evaluated with N = 100 and K = 2 using mixture of experts (Mo E), generalized product of experts (g Po E) and robust Bayesian committee machine (r BCM) aggregations and conditional probabilities (12). All GPs use an ARD squared exponential kernel and the hyperparameters are fitted using the first 1000 training samples for all methods. |