As a researcher in Deep Reinforcement Learning, I’m excited to share my latest project: Ano, a new optimizer designed to tackle noisy and non-convex environments. The idea is simple yet powerful: decouple the magnitude of the gradient from the direction of the momentum, making training more stable and faster. I’ve written a preprint and would love to get feedback from the community on both the method and the clarity of the writing.
Ano is available on Zenodo, and you can install it via pip. I’ve also set up a GitHub repository for experiments. This is my first real research contribution, so I’d greatly appreciate any feedback, suggestions, or constructive criticism.
As an independent researcher, I’m not affiliated with an institution, which means I need an endorsement to make the preprint available on arXiv. If anyone feels comfortable endorsing it after reviewing the paper, it would mean a lot.
Thanks for reading, and I look forward to hearing your thoughts!