对spikformer V2的疑问

[zxt6174]

对spikformer V2的疑问

[zxt6174]

作者您好。我对你们团队的工作很感兴趣。在spikformer v2中。为什么采用ANN作为Decoder？您在论文中只解释道用ANN的强大建模能力来提高SNN的表征能力。但Decoder自身的能力越强，对其输入的要求不就越低吗，也就是对Encoder输出的表征要求越低，这如何有利于Encoder的学习？恳请指教。

[zxt6174]

Hello author. I'm interested in your team's work. In spikformer v2. Why use ANN as a Decoder? In your paper, you only explain that the powerful modeling capabilities of ANNs can be used to improve the characterization of SNNs. However, the stronger the ability of the Decoder itself, the lower the requirements for its input, that is, the lower the representation requirements for the output of the Encoder, how is this conducive to the learning of the Encoder? Please advise.