Sensitivity-Guided Quantum Machine Unlearning
Post-training data removal from Variational Quantum Circuits
SQU is a novel quantum machine unlearning framework enabling post-training data removal from Variational Quantum Circuits (VQCs), addressing the Right to Be Forgotten (RTBF) requirements under GDPR.
Key Contributions
- Gradient-sensitivity analysis to identify and prune quantum circuit parameters encoding forgotten data
- Preserves model utility on retained data while efficiently unlearning target samples
- First quantum-native machine unlearning framework for VQCs