Novelty Detection Via Blurring
Authors: Sungik Choi, Sae-Young Chung
ICLR 2020 | Conference PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | 4 EXPERIMENTAL RESULTS |
| Researcher Affiliation | Academia | Sungik Choi & Sae-Young Chung School of Electrical Engineering Korea Advanced Institute of Science and Technology Daejeon, Republic of Korea {si_choi,schung}@kaist.ac.kr |
| Pseudocode | No | The paper does not contain structured pseudocode or algorithm blocks (clearly labeled algorithm sections or code-like formatted procedures). |
| Open Source Code | No | Footnote 1 states 'Our code is based on https://github.com/kuangliu/pytorch-cifar', but this does not unambiguously state that the code for the SVD-RND methodology described in this paper is made available or released by the authors at this link. It suggests using that repo as a base. |
| Open Datasets | Yes | We test novelty detection schemes on the blurred data generated by Singular Value Decomposition (SVD), we found that the novelty detection schemes assign higher confidence to the blurred data than the original data. ... VQ-VAE (Oord et al., 2017) in the CIFAR-10 (Krizhevsky et al., 2009) dataset. ... SVD-RND is examined in the cases in Table 1. CIFAR-10 : SVHN, Celeb A (Liu et al., 2015) : SVHN (Netzer et al., 2011), and Tiny Image Net (Deng et al., 2009) : (SVHN, CIFAR-10, CIFAR-100) are the cases studied by Nalisnick et al. (2019). We also studied CIFAR-10 : (LSUN (Yu et al., 2015), Tiny Image Net), LSUN : (SVHN, CIFAR-10, CIFAR-100) and Celeb A: (CIFAR-10, CIFAR-100) pairs. |
| Dataset Splits | Yes | For SVD-RND, we optimize the number of discarded singular values over different datasets. We choose the detector with the best performance across the validation data. ... The first 1000 images of the test OOD data are used for validation. |
| Hardware Specification | No | The paper does not provide specific hardware details (e.g., GPU models, CPU types, or memory specifications) used for running the experiments. |
| Software Dependencies | No | The paper states 'We implement the baselines and SVD-RND in the Py Torch framework.' and refers to 'The Adam optimizer', but it does not provide specific version numbers for PyTorch or other key software components. |
| Experiment Setup | Yes | Appendix C, titled 'DATA PREPROCESSING, NETWORK SETTINGS, PARAMETER SETTINGS FOR MAIN EXPERIMENT', provides specific details such as resizing images to 32x32, using Adam optimizer with a learning rate of 10^-4 annealed to 10^-5, specific epoch counts (50 or 34), and optimization ranges for parameters like K1, K2, and Gaussian kernel shapes. |