Cover First, Disagree Softly: Rethinking Mismatch-First Active Learning for Frame-Level Audio Classification

2026-07-15

AI Summary

This work addresses the high cost of strong-label annotation in frame-level audio classification and the limited performance of existing active learning methods under low labeling budgets, which often stems from neglecting sample similarity. To overcome these issues, the authors propose MW-FL, an active learning approach based on coverage prioritization and softened disagreement. MW-FL discards hard disagreement filtering and instead incorporates classifier disagreement as a hyperparameter-free weight into a facility location objective function, emphasizing coverage as the primary criterion. The method integrates geometric sampling with three disagreement-utilization strategies to construct a non-negative weighted coverage objective. Experiments on two multilabel datasets demonstrate that MW-FL significantly outperforms existing approaches such as MFFT, achieving state-of-the-art area under the learning curve, particularly under tight annotation budgets.

Abstract

Sound event detection relies on frame-level strong labels whose annotation is expensive. Active learning addresses this problem by selecting the audio segments whose labels help the classifier most. One of the prevailing acquisition strategies for this task, mismatch-first farthest-traversal (MFFT), combines the disagreement between two classifiers and the diversity of the selected segments through hard sequential decisions. It selects whole groups of high-disagreement segments first and spreads only the remaining budget by farthest traversal. On two multi-label datasets we show that this design is blind to the similarity among the selected segments and fails under low budgets, with every mismatch-first variant ending below the plain geometric strategy it builds on. We propose mismatch-weighted facility location (MW-FL), which spends the entire budget through a disagreement-weighted coverage objective that penalizes similarity among the selected segments. The disagreement signal from MFFT is used to obtain the nonnegative weights of this facility-location objective, without introducing hyperparameters. Experiments across two geometric mechanisms with three ways of using disagreement show that coverage of the selected segments is the dominant factor, hard disagreement gating of selection is harmful on both mechanisms, and soft disagreement weighting helps on top of coverage. MW-FL attains the best area under the learning curve on both datasets.

Problem

• active learning
• frame-level audio classification
• mismatch-first
• label efficiency
• sound event detection

Innovation

• active learning
• frame-level audio classification
• facility location
• disagreement weighting
• sample diversity