正则化 on 语音/音频论文速递 https://nanless.github.io/audio-paper-digest-blog/tags/%E6%AD%A3%E5%88%99%E5%8C%96/ Recent content in 正则化 on 语音/音频论文速递 Hugo zh-cn Wed, 29 Apr 2026 00:00:00 +0000 Inverse-Hessian Regularization for Continual Learning in ASR https://nanless.github.io/audio-paper-digest-blog/posts/2026-04-29-inverse-hessian-regularization-for-continual/ Wed, 29 Apr 2026 00:00:00 +0000 https://nanless.github.io/audio-paper-digest-blog/posts/2026-04-29-inverse-hessian-regularization-for-continual/ 语音识别 | 7.5/10 Regularized Inverse Filter Design for Rigid Spherical Microphone Array Processing: Laplace- And Time-Domain Representations https://nanless.github.io/audio-paper-digest-blog/posts/2026-04-29-regularized-inverse-filter-design-for-rigid/ Wed, 29 Apr 2026 00:00:00 +0000 https://nanless.github.io/audio-paper-digest-blog/posts/2026-04-29-regularized-inverse-filter-design-for-rigid/ 空间音频 | 8.0/10 Explicit Dropout: Deterministic Regularization for Transformer Architectures https://nanless.github.io/audio-paper-digest-blog/posts/2026-04-23-explicit-dropout-deterministic-regularization-for/ Thu, 23 Apr 2026 00:00:00 +0000 https://nanless.github.io/audio-paper-digest-blog/posts/2026-04-23-explicit-dropout-deterministic-regularization-for/ 这篇论文旨在解决传统Dropout方法依赖随机掩码、正则化效果不透明且难以精确控制的问题。其核心方法是提出一种确定性公式，将Dropout重新表述为一个可直接加入训练损失函数的显式正则化项，并推导出了适用于Transformer架构中注意力机制（Q、K、V）和前馈网络的正则化表达式。与已有方法相比，