https://steven-yl.github.io/mywebsite/tags/xavier/Xavier - 标签 - mywebsiteHugo -- gohugo.iozh-CNsteven@gmail.com (Steven)steven@gmail.com (Steven)This work is licensed under a Creative Commons Attribution-NonCommercial 4.0 International License.Fri, 03 Apr 2026 00:00:00 +0800https://steven-yl.github.io/mywebsite/net_init/Thu, 12 Mar 2026 00:00:00 +0800Stevenhttps://github.com/steven-ylhttps://steven-yl.github.io/mywebsite/net_init/全面对比深度学习权重初始化方法的原理、公式推导、优缺点与适用场景，附 PyTorch 代码示例和 Transformer 架构初始化最佳实践。https://steven-yl.github.io/mywebsite/kaiming/Fri, 03 Apr 2026 00:00:00 +0800Stevenhttps://github.com/steven-ylhttps://steven-yl.github.io/mywebsite/kaiming/用前向方差分析解释为何 ReLU 网络宜用方差 2/fan_in 的权重初始化，并对比 Xavier、给出 PyTorch 中的对应实现。