Graph-Guided Network for Irregularly Sampled Multivariate Time Series
Authors: Xiang Zhang, Marko Zeman, Theodoros Tsiligkaridis, Marinka Zitnik
ICLR 2022 | Conference PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | We use RAINDROP to classify time series and interpret temporal dynamics on three healthcare and human activity datasets. RAINDROP outperforms state-of-the-art methods by up to 11.4% (absolute F1-score points), including techniques that deal with irregular sampling using fixed discretization and set functions. RAINDROP shows superiority in diverse setups, including challenging leave-sensor-out settings. |
| Researcher Affiliation | Collaboration | Xiang Zhang Harvard University xiang_zhang@hms.harvard.edu Marko Zeman University of Ljubljana marko.zeman@fri.uni-lj.si Theodoros Tsiligkaridis MIT Lincoln Laboratory ttsili@ll.mit.edu Marinka Zitnik Harvard University marinka@hms.harvard.edu. M.Z. is supported, in part, by NSF under nos. IIS-2030459 and IIS-2033384, Harvard Data Science Initiative, Amazon Research Award, Bayer Early Excellence in Science Award, Astra Zeneca Research, and Roche Alliance with Distinguished Scientists Award. |
| Pseudocode | No | The paper does not contain any structured pseudocode or algorithm blocks. |
| Open Source Code | Yes | Code and datasets are available at https://github.com/mims-harvard/Raindrop. |
| Open Datasets | Yes | For all datasets used in this work, we share downloadable links to the raw sources and processed and ready-to-run datasets with the research community through this link: https://github.com/mims-harvard/Raindrop. |
| Dataset Splits | Yes | We randomly split the dataset into training (80%), validation (10%), and test (10%) set. The indices of these splits are fixed across all methods. |
| Hardware Specification | No | The paper does not provide specific hardware details (e.g., GPU/CPU models, memory) used for running experiments. |
| Software Dependencies | No | The paper mentions 'Python implementation' and 'Cytoscape' but does not provide specific version numbers for these or any other software dependencies. |
| Experiment Setup | Yes | In the generation of observation embedding, we set Ru as a 4-dimensional vector... The dimensions of time representation pt and rv are both 16. The trainable weight matrix D has shape of 4 32... We set the number of RAINDROP layers L as 2... We set the proportion of edge pruning as 50% (K=50)... The dk is set to 20, while the shape of W is 20 20... The da is set equal to the number of sensors. The first layer of ϕ has 128 neurons while the second layer has C neurons... We set λ = 0.02 to adjust Lr regularization scale. |