Multi-Head Attention with Disagreement Regularization
This work addresses a specific bottleneck in neural machine translation for researchers and practitioners, offering an incremental improvement to enhance model effectiveness.
The paper tackled the problem of insufficient diversity among attention heads in multi-head attention models by introducing disagreement regularization to encourage differences in subspaces, attended positions, and output representations, resulting in improved performance on WMT14 English-German and WMT17 Chinese-English translation tasks.
Multi-head attention is appealing for the ability to jointly attend to information from different representation subspaces at different positions. In this work, we introduce a disagreement regularization to explicitly encourage the diversity among multiple attention heads. Specifically, we propose three types of disagreement regularization, which respectively encourage the subspace, the attended positions, and the output representation associated with each attention head to be different from other heads. Experimental results on widely-used WMT14 English-German and WMT17 Chinese-English translation tasks demonstrate the effectiveness and universality of the proposed approach.