Invariance Calibration

How it works

Risks & contraindications

Highlighted claims

• The calibration step minimizes KL divergence between predictions on clean inputs and augmented views. — Asynchronous Federated Unlearning with Invariance Calibration for Medical Imaging
• Calibration is intended to penalize sensitivity to non-causal variations and reduce the chance that erasure is undone by later retained-client updates. — Asynchronous Federated Unlearning with Invariance Calibration for Medical Imaging
• Projected gradient ascent alone can suppress target behavior while leaving the model structurally close to a region where target features re-emerge. — Asynchronous Federated Unlearning with Invariance Calibration for Medical Imaging
• Moderate calibration produced low backdoor accuracy and high clean accuracy, while excessive calibration harmed clean accuracy. — Asynchronous Federated Unlearning with Invariance Calibration for Medical Imaging