🏢 Indian Institute of Science
Mitigating Biases in Blackbox Feature Extractors for Image Classification Tasks
·2583 words·13 mins·
loading
·
loading
Computer Vision
Image Classification
🏢 Indian Institute of Science
Researchers propose a simple yet effective clustering-based adaptive margin loss to mitigate biases inherited by black-box feature extractors in image classification tasks.
DisCEdit: Model Editing by Identifying Discriminative Components
·2619 words·13 mins·
loading
·
loading
Machine Learning
Deep Learning
🏢 Indian Institute of Science
DISCEDIT efficiently identifies and edits discriminative neural network components for structured pruning and class unlearning, achieving high sparsity and forgetting rates without needing training da…