ICML2025年时间检验奖（Test of Time）颁给了Batch Normalization。在这篇发表于2015年的论文中，作者提出深度神经网络训练中的“内部协变量偏移”问题。直观理解，就是隐藏层的数据分布会随着训练的进行而变化，而前一层的变化又会影响下一层的学习。这种层与层 ...

Normalization techniques have become integral to the training of deep neural networks, serving to stabilise learning dynamics, accelerate convergence and improve generality. At their core, these ...

AI training and inference are all about running data through models — typically to make some kind of decision. But the paths that the calculations take aren’t always straightforward, and as a model ...