发表机构
Uppsala University; Scilifelab（乌普萨拉大学; 瑞典生命科学实验室）
机构由 AI 辅助整理，请以论文原文为准。AI 中文总结
本研究比较八种分词策略于ECG Transformer，发现生理感知分词提升宏AUC 8.2%，并大幅减少序列长度和内存，无需增加模型容量。
AI 中文摘要
分词决定了呈现给ECG Transformer的生理内容以及注意力操作的序列。我们在九标签CPSC2018分类任务上，比较了Transformer、Informer、Reformer和FEDformer四种骨干网络下的八种分词策略。通过控制输入投影和主骨干网络容量，以隔离分词构建的影响。中位心跳和HeartLang分词在四种骨干网络上分别实现了0.893和0.889的平均宏AUC，而逐点和逐块分词分别为0.822和0.824。将两种生理感知表示进行池化，宏AUC相对提升了8.2%。同时，它们将平均序列长度从1,250个分词减少到158个，平均峰值训练内存从5.21 GB降至0.27 GB。结果表明，将分词与ECG形态对齐可以在不增加骨干网络容量的情况下，提高预测性能和内存效率。源代码可在该https URL上获取。
英文摘要
Tokenization determines both the physiological content presented to an ECG Transformer and the sequence over which attention operates. We compare eight tokenization strategies across Transformer, Informer, Reformer, and FEDformer on the nine-label CPSC2018 classification task. The input projection and principal backbone capacity are controlled to isolate the effect of token construction. Median-beat and HeartLang tokenization achieve mean macro-AUCs of 0.893 and 0.889 across the four backbones, compared with 0.822 and 0.824 for point-wise and patch-wise tokenization. Pooling the two physiology-aware representations yields an 8.2% relative improvement in macro-AUC. They also reduce mean sequence length from 1,250 to 158 tokens and mean peak training memory from 5.21 to 0.27 GB. The results show that aligning tokens with ECG morphology can improve both predictive performance and memory efficiency without increasing backbone capacity. The source code is available on https://github.com/LeeJarvis996/ecg_tokenizer.
