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..
Multiresolution Analysis and Statistical Thresholding on Dynamic Networks
Authors: Raphael Romero, Tijl De Bie, Nick Heard, Alexander Modell
NeurIPS 2025 | Venue PDF | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | In Section 5, we evaluate our method on both synthetic and real-world datasets, demonstrating that ANIE outperforms fixed-resolution approaches by effectively capturing changes at multiple time scales in dynamic networks. The proposed ANIE method is evaluated on two tasks. First, we generate synthetic Erd os-Renyi (ER) and Stochastic Block Model (SBM) datasets, and measure the performance of our method in estimating a known network intensity from an observed dynamic network. Second, in order to demonstrate the practical utility of our method, we apply it to the task of detecting change points in a real-world dataset of message interactions, and compare our method with two existing methods: Laplacian Anomaly Detection (LAD) [20] and Tensorsplat [26]. |
| Researcher Affiliation | Academia | Raphaël Romero Ghent University EMAIL Tijl De Bie Ghent University EMAIL Nick Heard Imperial College London EMAIL Alexander Modell Imperial College London EMAIL |
| Pseudocode | Yes | A full algorithmic description of the procedure is provided in the appendix. (Referring to Algorithm 1: Adaptive Network Intensity Estimation in Appendix A). |
| Open Source Code | Yes | An open-source implementation of the method is available at https://github.com/aida-ugent/anie. |
| Open Datasets | Yes | Furthermore, applications to real-world data showcase the practical benefits of ANIE s multiresolution approach to detecting structural change over fixed resolution methods. ... To this end, we apply ANIE to the UCI Messages dataset, as done in previous work [20]. ... The London Bike dataset, published by Transport for London [1], has an inherent dynamic network structure which has been previously studied for instance in [36, 41]. ... For the Enron dataset, as shown on Figure 8, we find that both the naive and reconstructed scaleograms capture changes in 2001, which marked the buildup to the company s bankruptcy. (Reference [1]: Transport for london cycle hire data. https://cycling.data.tfl.gov.uk/.) |
| Dataset Splits | No | The paper discusses generating synthetic datasets and applying ANIE to real-world datasets (UCI Messages, London Bike, Enron) for anomaly detection and case studies. However, it does not specify any training, testing, or validation splits for these datasets. For synthetic data, it mentions "mean and standard error over 10 runs" but no explicit data partitioning. |
| Hardware Specification | Yes | All our experiments were run on a Mac Book Air with a M1 chip and 8GB of RAM. (Also in Appendix D.3: Hardware used for the experiments All the experiments we run on a Mac Book Air with an Apple M1 chip with 8 CPU cores and 8GB of RAM.) |
| Software Dependencies | No | The paper mentions using 'Sci Py s sparse SVD implementation' but does not provide specific version numbers for any software dependencies. |
| Experiment Setup | Yes | Table 1 in Appendix D.2 provides hyperparameter selection for different methods: IPP-KDE Bandwidth (0.005, 0.05), IPP-Hist Number of bins (M) (128, 64), ANIE (ours) Resolution level (J) (8, 6), and Significance level (α) (0.05, 0.05). |