GAIA: Delving into Gradient-based Attribution Abnormality for Out-of-distribution Detection

Authors: Jinggang Chen, Junjie Li, Xiaoyang Qu, Jianzong Wang, Jiguang Wan, Jing Xiao

NeurIPS 2023 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental The effectiveness of GAIA is validated on both commonly utilized (CIFAR) and large-scale (Image Net-1K) benchmarks.
Researcher Affiliation Collaboration Huazhong University of Science and Technology, China Ping An Technology (Shenzhen) Co., Ltd. {chen.jinggang98, 2216217669ljj, quxiaoy}@gmail.com, jzwang@188.com, jgwan@hust.edu.cn, xiaojing661@pingan.com.cn
Pseudocode Yes Algorithm 1: GAIA
Open Source Code Yes Code is available at https://github.com/JGEthanChen/GAIA-OOD.
Open Datasets Yes The effectiveness of GAIA is validated on both commonly utilized (CIFAR) and large-scale (Image Net-1K) benchmarks.
Dataset Splits No The paper mentions using pre-trained models and various ID/OOD datasets for evaluation but does not specify the train/validation/test splits of these datasets used for training or validating their own models or experimental setup within the paper's text.
Hardware Specification No The paper mentions using ResNet34 and Wide Resnet40 models, but it does not specify the hardware (e.g., specific CPU, GPU models, memory, or computing cluster details) used for running the experiments.
Software Dependencies No The paper does not explicitly list specific software dependencies with version numbers (e.g., Python version, PyTorch version, specific library versions) that would be needed for reproducibility.
Experiment Setup No The paper states models are 'pre-trained with cross-entropy loss' and evaluates on benchmarks, but it does not provide specific hyperparameter values (e.g., learning rate, batch size, epochs, optimizer details) or detailed training configurations used for their experiments.