Notice: The reproducibility variables underlying each score are classified using an automated LLM-based pipeline, validated against a manually labeled dataset. LLM-based classification introduces uncertainty and potential bias; scores should be interpreted as estimates. Full accuracy metrics and methodology are described in Coakley et alK. L. Coakley, T. Snelleman, H. Hoos, and O. E. Gundersen, "The embrace of open science: An analysis of a decade of AI research and 56 800 conference papers," Under Review, 2026..
ScatterAD: Temporal-Topological Scattering Mechanism for Time Series Anomaly Detection
Authors: Tao Yin, Shaochen Fu, Zhibin Zhang, Li Huang, Xiaohong Zhang, Yiyuan Yang, Kaixiang Yang, Meng Yan
NeurIPS 2025 | Venue PDF | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | Extensive experiments on multiple public benchmarks show that Scatter AD achieves state-of-the-art performance on multivariate time series anomaly detection. We conducted evaluations on six real-world multivariate time series datasets, including (1) PSM (Pooled Server Metrics)[Abdulaal et al., 2021]; (2) MSL (Mars Science Laboratory)[Hundman et al., 2018]; (3) SWa T(Secure Water Treatment)[Mathur and Tippenhauer, 2016]; (4) WADI (Water Distribution)[Ahmed et al., 2017]; (5) NIPS-TS-SWAN[Lai et al., 2021]; (6) NIPS-TS-GECCO[Lai et al., 2021]. |
| Researcher Affiliation | Academia | 1Chongqing University 2University of Oxford 3South China University of Technology EMAIL, EMAIL EMAIL, EMAIL |
| Pseudocode | No | The paper describes the methodology but does not include any explicitly labeled 'Pseudocode' or 'Algorithm' block, nor does it present structured, step-by-step procedures formatted like code within the main text or figures. |
| Open Source Code | Yes | Code is available at this repository: https://github.com/jk-sounds/Scatter AD. |
| Open Datasets | Yes | We conducted evaluations on six real-world multivariate time series datasets, including (1) PSM (Pooled Server Metrics)[Abdulaal et al., 2021]; (2) MSL (Mars Science Laboratory)[Hundman et al., 2018]; (3) SWa T(Secure Water Treatment)[Mathur and Tippenhauer, 2016]; (4) WADI (Water Distribution)[Ahmed et al., 2017]; (5) NIPS-TS-SWAN[Lai et al., 2021]; (6) NIPS-TS-GECCO[Lai et al., 2021]. |
| Dataset Splits | Yes | For hyperparameter tuning, the training set was temporarily divided into 80% for training and 20% for validation. Table 3: Dataset statistics used in our experiments. Dims denotes the number of dimensions (variables). Train and Test indicate the number of time points in training and labeled test sets, respectively. |
| Hardware Specification | Yes | All experiments were performed in Python 3.9 using Py Torch and on NVIDIA Tesla-A800 GPUs. |
| Software Dependencies | No | All experiments were performed in Python 3.9 using Py Torch and on NVIDIA Tesla-A800 GPUs. The paper mentions 'Python 3.9' but does not specify a version number for 'Py Torch' or other key libraries, which is required for a reproducible description of ancillary software. |
| Experiment Setup | Yes | Initially, the batch size was set to 128, and the time window size was uniformly set to 110 for all datasets except NIPS-TS-GECCO(N-T-W) and NIP-TS-Swan(NT-S), which were set to 90. The GAT in Scatter AD has H=4 attention heads per layer and a hidden dimension of 512. For hyperparameter tuning, the training set was temporarily divided into 80% for training and 20% for validation. For each dataset, the learning rate was uniformly set to 0.0001. |