On the Properties of Kullback-Leibler Divergence Between Multivariate Gaussian Distributions

Authors: Yufeng Zhang, Jialu Pan, Li Ken Li, Wanwei Liu, Zhenbang Chen, Xinwang Liu, J Wang

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

Reproducibility Variable Result LLM Response
Research Type Experimental Finally, we discuss several applications of our theories in deep learning, reinforcement learning, and sample complexity research... We conduct extensive experiments to compare our method with 13 baseline methods... Table 2: Group-wise anomaly detection Results (AUROC and AUPR in percentage) of our method KLODS and the SOTA method GOD2KS on Glow with batch sizes 5 and 10. We run our method for 5 times. Results of GOD2KS are referred from [28].
Researcher Affiliation Academia Yufeng Zhang, Jialu Pan , Kenli Li College of Computer Science and Electronic Engineering Hunan University, Changsha, China {yufengzhang, jialupan, lkl}@hnu.edu.cn Wanwei Liu, Zhenbang Chen, Xinwang Liu College of Computer Key Laboratory of Software Engineering for Complex Systems National University of Defense Technology, Changsha, China {wwliu, zbchen, xinwangliu, wj}@nudt.edu.cn Ji Wang College of Computer State Key Laboratory for High Performance Computing Key laboratory of Software Engineering for Complex Systems National University of Defense Technology, Changsha, China wj@nudt.edu.cn
Pseudocode No The paper does not contain structured pseudocode or algorithm blocks.
Open Source Code No The paper discusses applications of its theoretical findings in a related work [57] but does not explicitly state that source code for the theoretical work or the specific KLODS algorithm mentioned is made publicly available within this paper or via a direct link.
Open Datasets Yes For example, Glow [29] assigns higher likelihoods for SVHN when trained on CIFAR-10... Table 2: Group-wise anomaly detection Results (AUROC and AUPR in percentage) of our method KLODS and the SOTA method GOD2KS on Glow with batch sizes 5 and 10. We run our method for 5 times. Results of GOD2KS are referred from [28]. ID OOD Fashion MNIST MNIST... CIFAR-10 Celeb A... CIFAR-10 CIFAR-100 SVHN LSUN
Dataset Splits No The paper mentions using In-Distribution (ID) and Out-of-Distribution (OOD) datasets for evaluation but does not provide specific train/validation/test dataset splits, percentages, or methodology for partitioning the data in this paper.
Hardware Specification No The paper does not provide specific hardware details (e.g., GPU/CPU models, processor types, or memory amounts) used for running the experiments mentioned in the application section.
Software Dependencies No The paper does not provide specific ancillary software details, such as library names with version numbers, needed to replicate the experiments.
Experiment Setup Yes Table 2: Group-wise anomaly detection Results (AUROC and AUPR in percentage) of our method KLODS and the SOTA method GOD2KS on Glow with batch sizes 5 and 10. We run our method for 5 times.