Track papers across labs, topics, and trends.
Search papers, labs, and topics across Lattice.
This paper introduces a novel two-phase transductive method for few-shot open-set audio classification that enhances prototype estimation by leveraging latent inlierness scores to minimize the influence of unknown-class samples. By refining prototypes primarily based on known-class evidence and employing a transductive loss that incorporates inlierness-weighted entropy measures, the approach effectively distinguishes between known and unknown classes. Experimental results demonstrate that this method achieves state-of-the-art performance across three audio datasets, addressing a critical challenge in open-set classification scenarios.
By effectively down-weighting potential unknown-class samples, this method achieves unprecedented accuracy in few-shot open-set audio classification.
Few-shot Open-set audio classification requires classifying query samples from known classes with a few labeled support samples while rejecting query samples from unknown classes. Transductive inference jointly observes the full unlabeled query set to improve prototype estimation, yet standard transductive updates do not distinguish known from unknown query samples, leaving prototypes vulnerable to open-set contamination. Drawing on latent-inlierness weighting and decoupled scoring for unknown-class samples, we propose a two-phase transductive method operating over a frozen audio encoder. First, each query sample is assigned a latent inlierness score that down-weights likely unknown-class samples, so that prototype refinement is driven primarily by known-class evidence. The refined prototypes are then directly optimized on a transductive loss combining support cross-entropy, inlierness-weighted conditional entropy minimization, and inlierness-weighted marginal entropy maximization, while open-set rejection uses a prior-adaptive free-energy score that adjusts its threshold with the prior proportion of unknown-class samples, decoupling detection from classification. Experiments on three audio datasets show our method achieves state-of-the-art results for few-shot open-set audio classification under multiple experimental conditions.