I am trying to understand how batch normalization (BN) works in CNNs. Suppose I have a feature map tensor $T$ of shape $(N, C, H, W)$

where $N$ is the mini-batch size,

$C$ is the number of channels, and

$H,W$ is the spatial dimension of the tensor.

Then it seems there could a few ways of going about this (where $\gamma, \beta$ are learned parameters for each BN layer)

Method 1: $T_{n,c,x,y} := \gamma*\frac {T_{c,x,y} - \mu_{x,y}} {\sqrt{\sigma^2_{x,y} + \epsilon}} + \beta$ where $\mu_{x,y} = \frac{1}{NC}\sum_{n, c} T_{n,c,x,y}$ is the mean for all channels $c$ for each batch element $n$ at spatial location $x,y$ over the minibatch, and

$\sigma^2_{x,y} = \frac{1}{NC} \sum_{n, c} (T_{n, c,x,y}-\mu_{c})^2$ is the variance of the minibatch for all channels $c$ at spatial location $x,y$.

Method 2: $T_{n,c,x,y} := \gamma*\frac {T_{c,x,y} - \mu_{c,x,y}} {\sqrt{\sigma^2_{c,x,y} + \epsilon}} + \beta$ where $\mu_{c,x,y} = \frac{1}{N}\sum_{n} T_{n,c,x,y}$ is the mean for a specific channel $c$ for each batch element $n$ at spatial location $x,y$ over the minibatch, and

$\sigma^2_{c,x,y} = \frac{1}{N} \sum_{n} (T_{n, c,x,y}-\mu_{c})^2$ is the variance of the minibatch for a channel $c$ at spatial location $x,y$.

Method 3: For each channel $c$ we compute the mean/variance over the entire spatial values for $x,y$ and apply the formula as

$T_{n, c,x,y} := \gamma*\frac {T_{n, c,x,y} - \mu_{c}} {\sqrt{\sigma^2_{c} + \epsilon}} + \beta$, where now $\mu_c = \frac{1}{NHW} \sum_{n,x,y} T_{n,c,x,y}$ and $\sigma^2{_c} = \frac{1}{NHW} \sum_{n,x,y} (T_{n,c,x,y}-\mu_c)^2 $

In practice which of these methods is used (if any) are correct for?

The original paper on batch normalization , https://arxiv.org/pdf/1502.03167.pdf , states on page 5 section 3.2, last paragraph, left side of the page:

For convolutional layers, we additionally want the normalization to obey the convolutional property – so that different elements of the same feature map, at different locations, are normalized in the same way. To achieve this, we jointly normalize all the activations in a minibatch, over all locations. In Alg. 1, we let $\mathcal{B}$ be the set of all values in a feature map across both the elements of a mini-batch and spatial locations – so for a mini-batch of size $m$ and feature maps of size $p \times q$, we use the effective mini-batch of size $m^\prime = \vert \mathcal{B} \vert = m \cdot pq$. We learn a pair of parameters $\gamma^{(k)}$ and $\beta^{(k)}$ per feature map, rather than per activation. Alg. 2 is modified similarly, so that during inference the BN transform applies the same linear transformation to each activation in a given feature map.

I'm not sure what the authors mean by "per feature map", does this mean per channel?


1 Answer 1


Method 2:
This is original batch Normalization as suggested in the paper [Ioffe & Szegedy, 2015].
It is the most common approach. It is very well explained here [d2l.ai]

Similarly, with convolutional layers, we can apply batch normalization after the convolution and before the nonlinear activation function. When the convolution has multiple output channels, we need to carry out batch normalization for each of the outputs of these channels, and each channel has its own scale and shift parameters, both of which are scalars. Assume that our minibatches contain m examples and that for each channel, the output of the convolution has height p and width q . For convolutional layers, we carry out each batch normalization over the m⋅p⋅q elements per output channel simultaneously. Thus, we collect the values over all spatial locations when computing the mean and variance and consequently apply the same mean and variance within a given channel to normalize the value at each spatial location

I'm not sure what the authors mean by "per feature map", does this mean per channel?

Yes, two trainable parameters per Channel/Feature map.

Method 3:
This was the idea suggested as "Layer Normalization". [Paper]
It fixed the issue of BN i.e. dependence on a Batch of data and also it worked for sequence data. But the paper didn't claim anything great for CNN.

We have also experimented with convolutional neural networks. In our preliminary experiments, we observed that layer normalization offers a speedup over the baseline model without normalization, but batch normalization outperforms the other methods. With fully connected layers, all the hidden units in a layer tend to make similar contributions to the final prediction and re-centering and rescaling the summed inputs to a layer works well. However, the assumption of similar contributions is no longer true for convolutional neural networks. The large number of the hidden units whose receptive fields lie near the boundary of the image are rarely turned on and thus have very different statistics from the rest of the hidden units within the same layer. We think further research is needed to make layer normalization work well in ConvNets

Method 1:
This is averaging across the Feature Maps on every pixel. I am not sure of its aplication.

  • $\begingroup$ If I'm not mistaken, your description from d2l.ai follows method 3, i.e. take for each channel the mean of the entire set of spatial locations summed and averaged for each minibatch element. My methods 1 and 2 were done for each individual spatial location separately, i.e they would have their own means and variances (1 being computed for all channels, 2 being the most specific, a mean and variance at each spatial location and at each channel). $\endgroup$ Commented Jan 21, 2021 at 15:42
  • $\begingroup$ Either way I understand how it works now. $\endgroup$ Commented Jan 21, 2021 at 15:44
  • $\begingroup$ No, scale and shift parameter is not the mean and std. This is also the answer to the new sub-question you have added. I will update. Though my logic to merge 1/2 seems inappropriate. Will update it too. $\endgroup$
    – 10xAI
    Commented Jan 21, 2021 at 16:31
  • $\begingroup$ I dont' think I said the scale and shift parameter were related to the mean or std, they are learned parameters. $\endgroup$ Commented Jan 21, 2021 at 20:26
  • $\begingroup$ If you look at my formulas for method 1,2 they are done for specific spatial locations which is very different than averaging over the entire x,y range of the feature map values. $\endgroup$ Commented Jan 21, 2021 at 20:28

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.