原书出处
Foundations of Computer Vision (Antonio Torralba, Phillip Isola etc.)
Layernorm computes a mean and variance over elements of a datapoint xin, and will do so separately for each such datapoint in a batch. Batchnorm computes the mean and variance per channel over datapoints in a batch. If we have a batch stored in the tensor , then what layernorm does looks just like a “transpose” of what batchnorm does. Batchnorm standardizes each element of
原书定位详情
来源:a19c7933-4eda-4176-aebd-30fe7e1ad292
正文位置:extracted/Foundations of Computer Vision (Antonio Torralba, Phillip Isola etc.)/_epub-src/OPS/xhtml/c012.xhtml
EPUB SHA-256:60a7eebc43f5d9a675d776fed3736fbdaa791e55877023f2f8cba016a9b384f8