Transitivity Recovering Decompositions: Interpretable and Robust Fine-Grained Relationships
Abhra Chaudhuri, Massimiliano Mancini, Zeynep Akata, Anjan Dutta
Neural Information Processing Systems, NeurIPS


Recent advances in fine-grained representation learning have come from leveraging local-to-global (emergent) relationships for achieving state-of-the-art results. The relational representations relied upon by such methods, however, are abstract. We aim to deconstruct this abstraction by expressing such relationships as interpretable graphs over image views. We begin by theoretically showing that abstract relational representations are nothing but a way of recovering transitive relationships among local views. Based on this, we design Transitivity Recovering Decompositions (TRD), a graph-space search algorithm that identifies interpretable equivalents of abstract emergent relationships at both instance and class levels, and with no post-hoc computations. We additionally discover that TRD is provably robust to noisy views, with empirical evidence also supporting this finding. The latter allows TRD to perform at par or even better than the state-of-the-art, while being fully interpretable.

(c) 2023 Explainable Machine Learning Tübingen Impressum