SAP: Corrective Machine Unlearning with Scaled Activation Projection for Label Noise Robustness
This addresses the problem of costly retraining and performance degradation due to label noise for machine learning practitioners, offering an incremental improvement over existing noise robust training methods.
The paper tackles label noise in training data by introducing Scaled Activation Projection (SAP), a corrective machine unlearning algorithm that improves model generalization, achieving up to 6% improvement on CIFAR with 25% synthetic corruption and 2.31% on a Vision Transformer with real-world noise.
Label corruption, where training samples are mislabeled due to non-expert annotation or adversarial attacks, significantly degrades model performance. Acquiring large, perfectly labeled datasets is costly, and retraining models from scratch is computationally expensive. To address this, we introduce Scaled Activation Projection (SAP), a novel SVD (Singular Value Decomposition)-based corrective machine unlearning algorithm. SAP mitigates label noise by identifying a small subset of trusted samples using cross-entropy loss and projecting model weights onto a clean activation space estimated using SVD on these trusted samples. This process suppresses the noise introduced in activations due to the mislabeled samples. In our experiments, we demonstrate SAP's effectiveness on synthetic noise with different settings and real-world label noise. SAP applied to the CIFAR dataset with 25% synthetic corruption show upto 6% generalization improvements. Additionally, SAP can improve the generalization over noise robust training approaches on CIFAR dataset by ~3.2% on average. Further, we observe generalization improvements of 2.31% for a Vision Transformer model trained on naturally corrupted Clothing1M.