Scaling for Training Time and Post-hoc Out-of-distribution Detection Enhancement

Authors: Kai Xu, Rongyu Chen, Gianni Franchi, Angela Yao

ICLR 2024 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental We experiment with CIFAR10, CIFAR100 (Krizhevsky, 2009), and Image Net-1K (Deng et al., 2009) ID data sources. [...] Metrics. We evaluate with (1) FPR@95, which measures the false positive rate at a fixed true positive rate of 95%, where lower scores are better and (2) AUROC (Area under the ROC curve).
Researcher Affiliation Academia Kai Xu1, Rongyu Chen1, Gianni Franchi2, Angela Yao1 1National University of Singapore 2U2IS, ENSTA Paris, Institut polytechnique de Paris
Pseudocode No The paper does not contain any pseudocode or clearly labeled algorithm blocks.
Open Source Code Yes Our code and models are available at https://github.com/kai422/SCALE.
Open Datasets Yes We experiment with CIFAR10, CIFAR100 (Krizhevsky, 2009), and Image Net-1K (Deng et al., 2009) ID data sources. [...] We used SVHN (Netzer et al., 2011), i SUN (Xu et al., 2015), Places365 (Zhou et al., 2018), and Textures (Cimpoi et al., 2014) as OOD datasets.
Dataset Splits Yes The Open OOD benchmark includes improved hyperparameter selection with a dedicated OOD validation set to prevent overfitting to the testing set." and "Table 2 uses p = 0.85 for SCALE and ASH-S, which is verified on the validation set.
Hardware Specification No The paper does not provide specific details about the hardware used for experiments, such as GPU or CPU models.
Software Dependencies No We adopted the Res Net-50 (He et al., 2016) model architecture and obtained the pretrained network from the torchvision library." (No version specified for torchvision or other libraries).
Experiment Setup Yes For training, we fine-tuned the torchvision pretrained model with ISH for 10 epochs with a cosine annealing learning rate schedule initiated at 0.003 and a minimum of 0. We additionally observed that using a smaller weight decay value (5e-6) enhances OOD detection performance.