Learning in the Wild with Incremental Skeptical Gaussian Processes

Authors: Andrea Bontempelli, Stefano Teso, Fausto Giunchiglia, Andrea Passerini

IJCAI 2020 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental Our experiments on synthetic and real-world data show that, as a result, while the original formulation of skeptical learning produces over-confident models that can fail completely in the wild, ISGP works well at varying levels of noise and as new classes are observed.
Researcher Affiliation Academia Andrea Bontempelli1 , Stefano Teso1 , Fausto Giunchiglia1,2 and Andrea Passerini1 1University of Trento, Italy 2Jilin University, Changchun, China name.surname@unitn.it
Pseudocode Yes Algorithm 1 Pseudo-code of ISGP. Y0 is provided as input. All branches are stochastic, see the relevant equations.
Open Source Code Yes The code and experimental setup can be downloaded from: gitlab.com/abonte/incremental-skeptical-gp.
Open Datasets No The paper mentions using a synthetic dataset and a 'location prediction task introduced in [Zeni et al., 2019]', but it does not provide a direct link, DOI, or specific repository name for either dataset used in its experiments, nor does it explicitly state the datasets are publicly available with concrete access information for this paper.
Dataset Splits Yes All results are 10-fold cross validated.
Hardware Specification Yes The experiments were run on a computer with a 2.2 GHz processor and 16 Gi B of memory.
Software Dependencies No The paper mentions 'implemented ISGP using Python 3', but it does not specify any libraries or frameworks with version numbers (e.g., PyTorch 1.x, scikit-learn 0.x).
Experiment Setup Yes All GP learners used a squared exponential kernel with a length scale of 2 and ρ = 10 8, without any optimization. The number of trees of SRF was set to 100.