An open API service indexing awesome lists of open source software.

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

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