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 [1].
xPatch: Dual-Stream Time Series Forecasting with Exponential Seasonal-Trend Decomposition
Authors: Artyom Stitsyuk, Jaesik Choi
AAAI 2025 | Venue PDF | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | We conduct extensive experiments on nine real-world multivariate time series datasets, including Electricity Transform Temperature (ETTh1, ETTh2, ETTm1, ETTm2) (Zhou et al. 2021), Weather, Traffic, Electricity, Exchangerate, ILI (Wu et al. 2021), and Solar-energy (Lai et al. 2018). Evaluation Metrics. Following previous works, we use Mean Squared Error (MSE) and Mean Absolute Error (MAE) metrics to assess the performance. |
| Researcher Affiliation | Collaboration | 1Korea Advanced Institute of Science and Technology (KAIST), South Korea 2INEEJI, South Korea EMAIL |
| Pseudocode | No | The paper describes the proposed method using mathematical equations and textual explanations, but it does not contain explicit pseudocode blocks or algorithm listings. |
| Open Source Code | Yes | Code https://github.com/stitsyuk/x Patch |
| Open Datasets | Yes | We conduct extensive experiments on nine real-world multivariate time series datasets, including Electricity Transform Temperature (ETTh1, ETTh2, ETTm1, ETTm2) (Zhou et al. 2021), Weather, Traffic, Electricity, Exchangerate, ILI (Wu et al. 2021), and Solar-energy (Lai et al. 2018). |
| Dataset Splits | No | Unified Experimental Settings. To ensure a fair comparison, we conduct 2 types of experiments. The first experiment uses unified settings based on the forecasting protocol proposed by Times Net (Wu et al. 2023): a lookback length L = 36, prediction lengths T = {24, 36, 48, 60} for the ILI dataset, and L = 96, T = {96, 192, 336, 720} for all other datasets. The paper defines lookback and prediction lengths but does not explicitly state the train/validation/test split percentages for the datasets within this document. |
| Hardware Specification | Yes | All the experiments are implemented in Py Torch (Paszke et al. 2019), and conducted on a single Quadro RTX 6000 GPU. |
| Software Dependencies | No | All the experiments are implemented in Py Torch (Paszke et al. 2019). The paper mentions PyTorch but does not specify a version number for it or any other software dependencies. |
| Experiment Setup | Yes | Unified Experimental Settings. To ensure a fair comparison, we conduct 2 types of experiments. The first experiment uses unified settings based on the forecasting protocol proposed by Times Net (Wu et al. 2023): a lookback length L = 36, prediction lengths T = {24, 36, 48, 60} for the ILI dataset, and L = 96, T = {96, 192, 336, 720} for all other datasets. It is also important to note that we strictly adhere to the settings specified in the official implementations, including the number of epochs (100 for CARD and Patch TST, 15 for RLinear) and the learning rate adjustment strategy. |