Improving Bayesian Neural Networks by Adversarial Sampling

Authors: Jiaru Zhang, Yang Hua, Tao Song, Hao Wang, Zhengui Xue, Ruhui Ma, Haibing Guan10110-10117

AAAI 2022 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental We conduct extensive experiments with multiple network structures on different datasets, e.g., CIFAR-10 and CIFAR-100. Experimental results validate the correctness of the theoretical analysis and the effectiveness of the Adversarial Sampling on improving model performance.
Researcher Affiliation Academia 1 Shanghai Jiao Tong University 2 Queen s University Belfast 3 Louisiana State University
Pseudocode Yes Algorithm 1: Training with Adversarial Sampling
Open Source Code Yes We release our codes at https://github.com/AISIGSJTU/AS.
Open Datasets Yes We train a variety of Bayesian neural networks, including Res Net20, Res Net56 (He et al. 2016), and VGG (Simonyan and Zisserman 2015), on CIFAR-10, and CIFAR-100 datasets (Krizhevsky 2009).
Dataset Splits No The paper mentions training on CIFAR-10 and CIFAR-100 datasets and evaluating on test sets, but does not explicitly provide details about training/validation/test splits (e.g., percentages, sample counts, or specific cross-validation setup).
Hardware Specification No The paper does not provide specific details regarding the hardware (e.g., GPU models, CPU types, memory) used for running the experiments.
Software Dependencies No The paper does not specify software dependencies with version numbers (e.g., 'Python 3.8', 'PyTorch 1.9').
Experiment Setup Yes For simplicity, we set the hyperparameter α = 0.02 and N = 5 on models trained with Adversarial Sampling.