Concept activation vectors can be computed as the normalized difference between concept-mean and global-mean activations, giving a 46.4x average speedup over SVM-based CAVs with comparable quality.
Network dissection: Quantifying interpretability of deep visual representations
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.LG 1years
2025 1verdicts
CONDITIONAL 1representative citing papers
citing papers explorer
-
FastCAV: Efficient Computation of Concept Activation Vectors for Explaining Deep Neural Networks
Concept activation vectors can be computed as the normalized difference between concept-mean and global-mean activations, giving a 46.4x average speedup over SVM-based CAVs with comparable quality.