ChronoEpilogi: Scalable Time Series Selection with Multiple Solutions
Authors: Etienne Vareille, Michele Linardi, Ioannis Tsamardinos, Vassilis Christophides
NeurIPS 2024 | Conference PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | Experiments on synthetic and real datasets demonstrate the scalability of Chrono Epilogi to hundreds of TS variables and its efficacy in identifying multiple solutions. In the real datasets, Chrono Epilogi is shown to reduce the number of TS variables by 96% (on average) by conserving or even improving forecasting performance. |
| Researcher Affiliation | Academia | Etienne Vareille1 Michele Linardi1 Ioannis Tsamardinos2 Vassilis Christophides1 1ETIS UMR-8051 Laboratory, CY Cergy Paris Universite, ENSEA, CNRS 2Computer Science Department, University of Crete, Heraklion, Greece |
| Pseudocode | Yes | Algorithm 1 Chrono Epilogi-FBE Require: TS X, target T, max lag L, threshold params [α, γ, δ] 1: set S FORWARD(X, T, L,α) 2: S BACKWARD(X, T, L,S,γ) 3: set M EQUIV(X, T, L,S,δ) 4: return M set of eq. Markov bound. |
| Open Source Code | Yes | 2https://github.com/ev07/Chrono Epilogi |
| Open Datasets | Yes | Real datasets We evaluate our approach on five forecasting datasets covering different domains: Electricity (consumption), Solar (production), S.F. Traffic [GBW+21], METR-LA, and PEMS-BAY [LYSL18] (transport). |
| Dataset Splits | Yes | The Tuning set is itself separated into five folds along the time axis, to conduct hyperparameter optimization. |
| Hardware Specification | Yes | We run our experiments on servers runing Ubuntu 22.04.4 LTS, with 36 cores, 1TB RAM, GPU Quadro RTX 8000 with driver version 550.54.14 and CUDA version 12.4. |
| Software Dependencies | No | We include in the repository a requirements file requirements.txt listing the necessary dependencies. However, specific version numbers are not listed in the paper's text. |
| Experiment Setup | Yes | We optimize the hyperparameters for each algorithm and forecaster we consider (see Table 3). We use the python library optuna with Grid Search optimization. Chrono Epilogi thresholds cover several orders of value between 10 20 and 0.05. As we observed that for Group Lasso regularization parameter 10 20, all TS where select no matter the dataset, and at 0.1, no TS was selected other than the target past, Group Lasso group regularization parameter ranges 25 values within this range on a logarithmic scale. |