ParallelTime

KO | EN

gitlite — search

#python #machine-learning #rust #deep-learning #docker #java #windows #awesome #cli #kubernetes #pytorch #security #mcp #openai

(★ 20)

Official Implimitation of ParallelTime: Dynamically Weighting the Balance of Short- and Long-Term Temporal Dependencies. https://www.arxiv.org/abs/2507.13998

ParallelTime 최신버젼 다운로드

최종 버전 다운로드 (.zip)

// repository documentation

Was this content helpful?

★ 0 (0 ratings)

# <center>ParallelTime</center>

### Welcome to the official repository of [ParallelTime: Dynamically Weighting the Balance of Short- and Long-Term Temporal Dependencies](https://www.arxiv.org/abs/2507.13998). 
### The architecture of ParallelTime.

### Our main contributions:
* ParallelTime (Parallel Time) selects the contributions of short-term, long-term, and global memory for each time series patch.
* We demonstrate that parallel Mamba-Attention architecture is currently the most effective
approach for long-term time series forecasting.
* Our model, ParallelTime, achieves SOTA performance on real-world benchmarks, delivering better results from previews models with fewer parameters and lower computational cost, critical for real-time forecasting.
## Results
<img src="figures/results.png" alt="ParallelTime Architucture" width="100%">

## Usage of Parallel Time

1. Obtain the fundamental long-term forecasting datasets, including Weather, Illness, Traffic, Electricity, and ETT (comprising 4 datasets)
You can do it by downloading it from this [Autoformer](https://drive.google.com/drive/folders/1ZOYpTUa82_jCcxIdTmyr0LXQfvaM9vIy), and move them into `dataset/` directory.
```
├── dataset
│   ├── electricity
│   │   └── electricity.csv
│   ├── ETT-small
│   │   ├── ETTh1.csv
│   │   ├── ETTh2.csv
│   │   ├── ETTm1.csv
│   │   └── ETTm2.csv
│   ├── illness
│   │   └── national_illness.csv
│   ├── traffic
│   │   └── traffic.csv
│   └── weather
│       └── weather.csv 
```

2. Install requirements. `pip install uv` followed by `uv sync` which will install all the project dependencies (dont forget to change the source to .venv).

3. Look through our scripts located at ```./scripts``` which run 4 runs for diffrent prediction length depend on the dataset. You'll find the core of ParallelTime in ```models/ParallelTime.py```. For example, to get the multivariate forecasting results for ETTh1 dataset, just run the following command `sh ./scripts/etth1.sh`

For the other datasets run on a Linux machine:
```
sh ./scripts/<script_name>.sh
```
After the run ends for each prediction length, you can open `./result_long_term_forecast_<dataset_name>.txt` to view the results once the model run is complete. Additionally, logs and configurations are saved for each run in ```./multirun/current_date/<dataset_name>/run_time```, which can be used to review more detailed results.

## Acknowledgement

We are deeply grateful for the valuable code and efforts contributed by the following GitHub repositories.
- https://github.com/thuml/iTransformer
- https://github.com/yuqinie98/PatchTST

## Cite ParallelTime
If ParallelTime or this repository is useful in your own research, you can use the following BibTeX entry:
```
@misc{katav2025paralleltimedynamicallyweightingbalance,
      title={ParallelTime: Dynamically Weighting the Balance of Short- and Long-Term Temporal Dependencies}, 
      author={Itay Katav and Aryeh Kontorovich},
      year={2025},
      eprint={2507.13998},
      archivePrefix={arXiv},
      primaryClass={cs.LG},
      url={https://arxiv.org/abs/2507.13998}, 
}
```
## Contact
If you have any questions or concerns, please contact us or submit an issue

ParallelTime

ParallelTime 최신버젼 다운로드

오늘의 추천 프로젝트

Was this content helpful?

Recent Feedback

ParallelTime

ParallelTime 최신버젼 다운로드

오늘의 추천 프로젝트

Was this content helpful?

Recent Feedback

Download README