Robustness Guarantees for Bayesian Inference with Gaussian Processes

Authors: Luca Cardelli, Marta Kwiatkowska, Luca Laurenti, Andrea Patane7759-7768

AAAI 2019 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental We evaluate our techniques on two examples, a GP regression problem and a fully-connected deep neural network, where we rely on weak convergence to GPs to study adversarial examples on the MNIST dataset.
Researcher Affiliation Collaboration 1Microsoft Research Cambridge, 2University of Oxford
Pseudocode No The paper describes algorithmic methods, but it does not include a clearly labeled pseudocode block or algorithm steps in a structured format.
Open Source Code Yes Code available at: https://github.com/andreapatane/checkGP.
Open Datasets Yes we train a selection of Re LU GPs on a subset of the MNIST dataset
Dataset Splits No The paper discusses training on a subset of the MNIST dataset using least-square classification and notes "Classification accuracy obtained on the full MNIST test set". It mentions using 100 to 2000 training samples, but does not specify a validation split or how one would reproduce it.
Hardware Specification No The paper does not provide specific details about the hardware (e.g., GPU/CPU models, memory) used for running the experiments.
Software Dependencies No The paper does not provide specific software dependencies with version numbers (e.g., 'Python 3.8, PyTorch 1.9'). It mentions using SIFT and discusses GPs and NNs, but without versioned software details.
Experiment Setup Yes Unless otherwise stated, we perform analysis on the best model obtained using 1000 training samples, that is, a two-hidden-layer architecture with σ2 w = 3.19 and σ2 b = 0.00.