ConjNorm: Tractable Density Estimation for Out-of-Distribution Detection
Authors: Bo Peng, Yadan Luo, Yonggang Zhang, Yixuan Li, Zhen Fang
ICLR 2024 | Conference PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | Extensive experiments across OOD detection benchmarks empirically demonstrate that our proposed CONJNORM has established a new state-of-the-art in a variety of OOD detection setups, outperforming the current best method by up to 13.25% and 28.19% (FPR95) on CIFAR-100 and Image Net-1K, respectively. |
| Researcher Affiliation | Academia | Bo Peng1 , Yadan Luo2 , Yonggang Zhang3, Yixuan Li4, Zhen Fang1 University of Technology Sydney, Australia1 The University of Queensland, Australia2 Hong Kong Baptist University, Hong Kong3 University of Wisconsin-Madison, USA4 bo.peng-7@student.uts.edu.au, y.luo@uq.edu.au csygzhang@comp.hkbu.edu.hk, sharonli@cs.wisc.edu zhen.fang@uts.edu.au |
| Pseudocode | No | The paper describes methods and equations but does not include any explicitly labeled pseudocode or algorithm blocks. |
| Open Source Code | Yes | To make all experiments reproducible, we have listed all detailed hyper-parameters. We upload source codes and instructions in the supplementary materials. |
| Open Datasets | Yes | We consider CIFAR10 and CIFAR-100 (Krizhevsky et al., 2009) as ID data and train Dense Net-101 (Huang et al., 2017) on them respectively... We conduct experiments on the Image Net benchmark... OOD datasets include i Naturalist (Xiao et al., 2010b), SUN (Xiao et al., 2010a), Places365 (Zhou et al., 2017), and Textures (Cimpoi et al., 2014). |
| Dataset Splits | No | The paper describes training and test sets but does not explicitly provide details about specific validation splits (e.g., percentages or counts) or their usage in the main text. |
| Hardware Specification | No | The paper does not specify the exact GPU models, CPU types, or other detailed computer specifications used for running the experiments. It only mentions general models like DenseNet, ResNet, and MobileNet. |
| Software Dependencies | No | The paper mentions "Pytorch (Paszke et al., 2019)" but does not specify a version number for PyTorch or any other software dependencies, which is required for reproducibility. |
| Experiment Setup | Yes | For both CIFAR-10 and CIFAR-100, the model is trained for 100 epochs, with batch size 64, weight decay 1e-4, and Nesterov momentum 0.9. The start learning rate is 0.1 and decays by a factor of 10 at 50th, 75th, and 90th epochs. |