K1nght/Unified-Unlearning-w-Remain-Geometry
[NeurIPS2024 (Spotlight)] "Unified Gradient-Based Machine Unlearning with Remain Geometry Enhancement" by Zhehao Huang, Xinwen Cheng, JingHao Zheng, Haoran Wang, Zhengbao He, Tao Li, Xiaolin Huang
This project helps data scientists and machine learning engineers quickly and efficiently remove specific learned information from large-scale deep neural networks without retraining the entire model. It takes an existing trained model and a request to "forget" certain data (like specific images or classes), outputting a modified model that no longer contains that information. This is useful for maintaining data privacy or adapting models to new regulations.
Use this if you need to erase particular data or classes from an already deployed deep learning model, especially in computer vision tasks, without the time and computational cost of full retraining.
Not ideal if you need to entirely rebuild a model from scratch with new data, rather than selectively forgetting existing information, or if you are not working with deep neural networks.
Stars
12
Forks
—
Language
Python
License
—
Category
Last pushed
Oct 17, 2025
Commits (30d)
0
Get this data via API
curl "https://pt-edge.onrender.com/api/v1/quality/diffusion/K1nght/Unified-Unlearning-w-Remain-Geometry"
Open to everyone — 100 requests/day, no key needed. Get a free key for 1,000/day.
Higher-rated alternatives
OPTML-Group/Unlearn-Saliency
[ICLR24 (Spotlight)] "SalUn: Empowering Machine Unlearning via Gradient-based Weight Saliency in...
Shilin-LU/VINE
[ICLR 2025] "Robust Watermarking Using Generative Priors Against Image Editing: From...
WindVChen/DiffAttack
An unrestricted attack based on diffusion models that can achieve both good transferability and...
koninik/DiffusionPen
Official PyTorch Implementation of "DiffusionPen: Towards Controlling the Style of Handwritten...
Wuyxin/DISC
(ICML 2023) Discover and Cure: Concept-aware Mitigation of Spurious Correlation