Published September 15, 2025
| Version v5
Preprint
Open
Dynamic Fold Gradient Descent (DFGD):New AI Algorithm methodology
Creators
Description
Dynamic Fold Gradient Descent (DFGD):This optimization algorithm is a gradient-sampling-based approach that losslessly boosts efficiency by integrating dimension folding, pruning, and quantization. It offers significant potential for future development and can be applied to any gradient-based optimizer, such as Adam.The DOI for t's math theory:10.5281/zenodo.17115298
Files
v5-DFGD.pdf
Files
(402.3 kB)
| Name | Size | Download all |
|---|---|---|
|
md5:8bd947b57d0b6b6c680c4b9730be486f
|
402.3 kB | Preview Download |
Additional details
References
- Sulin, Z. (2025). Folding Convergence Criterion and Its Implications for Algorithm Optimization. Zenodo. https://doi.org/10.5281/zenodo.17115298