NECO: NEural Collapse Based Out-of-distribution detection
This addresses a critical challenge in ML for improving model reliability by preventing overconfidence on unseen data, representing a novel application of neural collapse rather than an incremental improvement.
The paper tackles the problem of detecting out-of-distribution (OOD) data in machine learning by introducing NECO, a post-hoc method based on neural collapse and principal component spaces, achieving state-of-the-art results on various OOD detection tasks with strong generalization across architectures.
Detecting out-of-distribution (OOD) data is a critical challenge in machine learning due to model overconfidence, often without awareness of their epistemological limits. We hypothesize that ``neural collapse'', a phenomenon affecting in-distribution data for models trained beyond loss convergence, also influences OOD data. To benefit from this interplay, we introduce NECO, a novel post-hoc method for OOD detection, which leverages the geometric properties of ``neural collapse'' and of principal component spaces to identify OOD data. Our extensive experiments demonstrate that NECO achieves state-of-the-art results on both small and large-scale OOD detection tasks while exhibiting strong generalization capabilities across different network architectures. Furthermore, we provide a theoretical explanation for the effectiveness of our method in OOD detection. Code is available at https://gitlab.com/drti/neco