arXivDaily arXiv每日学术速递 周一至周五更新
arXiv周末暂无论文更新,休息一下吧,周末愉快~~
arXiv 2403.13086cs.SDcs.LGeess.ASeess.SP
收藏

面向音频分类器的可听化映射(Listenable Maps for Audio Classifiers)

Listenable Maps for Audio Classifiers

Fondazione Bruno Kessler(布鲁诺·凯塞勒基金会);Mila-Québec AI Institute(米拉-魁北克人工智能研究所);Concordia University(康考迪亚大学);Laval University(拉瓦尔大学)

Francesco Paissan, Mirco Ravanelli, Cem Subakan

AI总结:

针对音频分类器解释难的问题,提出事后解释方法L-MAC,通过解码器生成二值掩码高亮音频相关部分,实验证明其解释忠实度优于现有方法且更受用户偏好。

摘要:

尽管深度学习模型在各类任务中表现亮眼,但其复杂性给可解释性带来了挑战。这一挑战在音频信号领域尤为突出,因为传递解释结果本身就存在固有难度。为解决该问题,我们提出了面向音频分类器的可听化映射(Listenable Maps for Audio Classifiers, L-MAC),这是一种事后解释方法,可生成忠实且可听的解释结果。L-MAC在预训练分类器的顶部接入一个解码器,用于生成二值掩码,以高亮显示输入音频中的相关部分。我们采用特定损失函数训练解码器:该损失函数最大化分类器对音频掩码保留部分的决策置信度,同时最小化模型对掩码移除部分的输出概率。在域内和域外数据上的定量评估表明,与多种基于梯度和掩码的方法相比,L-MAC始终能生成更忠实的解释结果。此外,用户研究证实,用户平均而言更偏好该技术生成的解释。

英文摘要:

Despite the impressive performance of deep learning models across diverse tasks, their complexity poses challenges for interpretation. This challenge is particularly evident for audio signals, where conveying interpretations becomes inherently difficult. To address this issue, we introduce Listenable Maps for Audio Classifiers (L-MAC), a posthoc interpretation method that generates faithful and listenable interpretations. L-MAC utilizes a decoder on top of a pretrained classifier to generate binary masks that highlight relevant portions of the input audio. We train the decoder with a loss function that maximizes the confidence of the classifier decision on the masked-in portion of the audio while minimizing the probability of model output for the masked-out portion. Quantitative evaluations on both in-domain and out-of-domain data demonstrate that L-MAC consistently produces more faithful interpretations than several gradient and masking-based methodologies. Furthermore, a user study confirms that, on average, users prefer the interpretations generated by the proposed technique.

机构、总结、中文摘要均由AI分析而来,可能存在错误,请以论文正文为准。

↑