https://github.com/timetoai/timediffusion_synth
Research Project on synthetic time-series generation
https://github.com/timetoai/timediffusion_synth
data-science deep-learning machine-learning pytorch synthetic-time-series time-series
Last synced: over 1 year ago
JSON representation
Research Project on synthetic time-series generation
- Host: GitHub
- URL: https://github.com/timetoai/timediffusion_synth
- Owner: timetoai
- Created: 2022-10-09T19:15:03.000Z (over 3 years ago)
- Default Branch: main
- Last Pushed: 2022-12-17T06:14:01.000Z (over 3 years ago)
- Last Synced: 2025-01-24T22:11:54.656Z (over 1 year ago)
- Topics: data-science, deep-learning, machine-learning, pytorch, synthetic-time-series, time-series
- Language: Jupyter Notebook
- Homepage:
- Size: 10.6 MB
- Stars: 4
- Watchers: 1
- Forks: 0
- Open Issues: 0
-
Metadata Files:
- Readme: README.md
Awesome Lists containing this project
README
# TimeDiffusion
Diffusion based temporal convolutional model for synthetic time-series generation. Main focus is on coherence of autoregressive models' results on real and synthetic data.
Project also includes several time-series generation models implementation for performance comparison:
- QuantGAN
- TTS GAN
- RealNVP
- FourierFlow
**Project structure**
* [results](./results) - folder with csv results files
* [utils](./utils)
* [utils/dl.py](./utils/dl.py) - time-series deep learning models in pytorch with some decorators for training / inference
* [utils/synth_eval.py](./utils/synth_eval.py) - functions for models evaluation
* [utils/timediffusion.py](./utils/timediffusion.py) - TimeDiffusion model
* [results](./results) - folder with csv results files
* _train_synth.ipynb - jupyter notebooks with training and subsequent generation of synthetic data for specific model
* [synth_model_evaluation.ipynb](./synth_model_evaluation.ipynb) - jupyter notebook with example of generation model quality evaluation
* [results_visualization.ipynb](./results_visualization.ipynb) - jupyter notebook with visualization of generation models comparison