Collect Ideas for the PyAF 5.0 roadmap

[antoinecarme]

This is an issue to serve as a honeypot for all ideas that come and can be included in the next PyAF release (5.0, expected on 2023-07-14).

First ideas :

1. Forecasting Models based on Deep Learning Attention Mechanisms
2. Recurrent tasks : More hardware architectures, forecasting performance measures, benchmarks, bug fixes, docs and optimization/profiling.
3. User feedback and issues

Ideally, the main part of the implementation will take place during the coming covid lockdown (Fall 2022 ;-).

[antoinecarme]

Large Horizon Models (H large enough). Profiling for CPU/memory/speed.

https://github.com/antoinecarme/pyaf/issues/213

[antoinecarme]

Model Esthetics

https://github.com/antoinecarme/pyaf/issues/212

[antoinecarme]

Get rid of Keras and tensorflow. Use PyTorch. Less is better. Minimal is beautiful.

https://github.com/antoinecarme/pyaf/issues/211

[antoinecarme]

PyAF will use PyTorch as its deep learning architecture for future projects.

https://github.com/antoinecarme/pyaf/issues/211

[antoinecarme]

Investigate business days / hours impact.

https://github.com/antoinecarme/pyaf/issues/210

[antoinecarme]

Add some outlier-resistant forecasting performance measures for robustness.

https://github.com/antoinecarme/pyaf/issues/209

[antoinecarme]

New Hardware Architecture : RISC-V

https://github.com/antoinecarme/pyaf/issues/208

[antoinecarme]

Investigate Threshold AR (TAR) Models

https://github.com/antoinecarme/pyaf/issues/214

[antoinecarme]

Investigate TSMARS Models

215

[antoinecarme]

221

[antoinecarme]

Implemented/Processed So far :

2. 208

3. 209

4. 211

5. 212

6. 213

7. 216

8. 217

9. 220

10. 221

11. 223

12. 224

13. 225

[antoinecarme]

Release Process starts.

228

CLOSING