Here are some links on machine unlearning – Part 1:
- Announcing the first Machine Unlearning Challenge – Google Research Blog (googleblog.com)
- Now That Machines Can Learn, Can They Unlearn? | WIRED
- Fact checks using free and open source information via search / encyclopedia and rerun the whole model – Specific usecase for hallucinations in GAI / LLM for specific prompt to get facts
- Retrain by excluding the content – Expensive, time consuming and can’t guarantee in case of input requests coming which are close to original as model might classify / work on them correctly
- Block the results and inputs which are infringing on privacy but again this is expensive and time consuming
- [2209.02299] A Survey of Machine Unlearning (arxiv.org)
- GitHub – tamlhp/awesome-machine-unlearning: Awesome Machine Unlearning (A Survey of Machine Unlearning)
- [1912.03817] Machine Unlearning (arxiv.org)
- GitHub – jjbrophy47/machine_unlearning: Existing Literature about Machine Unlearning
- Starting kit now available | NeurIPS 2023 Machine Unlearning Challenge (unlearning-challenge.github.io)
- Challenge is to find a fundamental, programming agnostic approach like clustering, ANN, CNN and so on especially for deep learning which can help with requests for privacy, right to forget and similar
Email me: Neil@HarwaniSystems.in