metadata
license: mit
datasets:
- mnist
metrics:
- accuracy
pipeline_tag: image-classification
A super tiny mnist model to demostrate the potential of the leanable non-linearity OptAEG-V1.
The model reaches 96.3% accuracy with only 1.5k parameter size.
The OptAEG-V1 leanable non-linearity is based on a thery of Arithmetic Expression Geometry. The theory is still in developing, please reference https://github.com/mountain/aeg-paper