A linear-time alternative to standard momentum-based optimizers for training neural networks. It uses unit-norm momentum updates to preserve structural alignment without the computational cost of semi-orthogonal matrices.
Tech Stack
PythonAI, ML & Data
Why AuON?
• Linear time complexity
• Better structural alignment
• No semi-orthogonal matrices
Limitations
• Niche optimizer
• Requires PyTorch integration
• Limited documentation
1/18/2026
Last Update
0
Forks
0
Issues
Apache-2.0
License
Financial Leak Detected
Stop the "SaaS Tax"
Your team could be burning cash. Switching to AuON instantly boosts your runway.
