Questions tagged [batch-normalization]
For questions about Batch Normalization of layer activations in theory and practice, as used in (typically deep) neural networks.
62
questions
0
votes
2
answers
47
views
Dropout and BatchNorm decrease speed of learning
Experimenting with the cifar10 dataset and faced with strange behavior when Dropout and BatchNorm don't help at all.
As I get:
Dropout - freezing some of the weights which helps us to prevent ...
- 103
0
votes
0
answers
13
views
Why do we use Gradient Descent for Norm Batch since the function has only 2 parameters?
So here is the formula for Z sedile ( gamma * Z + Beta). Since it's a function of 2 parameters why there is still stated in courses that we need to compute gradient descent to find the slope ? Since ...
0
votes
0
answers
15
views
Batch normalization getting Val accuracy of 000e00
I am currently stuck on batch normalization. I have code written out but when I do it it keeps giving me a val_accuracy of 00e00 and stops after 2 epochs (I am wanting to run 20). I am not sure what ...
1
vote
2
answers
156
views
Should we always use Batch Renormalization over Batch Normalization?
According to a machine level mastery post on batch norm
For small mini-batch sizes or mini-batches that do not contain a
representative distribution of examples from the training dataset, the
...
- 187
0
votes
0
answers
19
views
Batchnorm and gradient attenuation
Am I understood right that if we put batchnorm layer after every conv layer and our network is very deep, then we won't have a problem of gradient attenuation? In other words, is it true that ...
- 101
0
votes
0
answers
81
views
Tensorflow: ShuffleNetV2 Implementation
I'm trying to implement shufflenetv2 model by using tensorflow and keras, using the following code down below:
...
- 101
0
votes
0
answers
68
views
BatchNormalization not working as expected in a 1D convolutional net
I'm trying to make a neural net and have problems with normalization of the values of the hidden layers.
I've tried using a BatchNormalization layer to fix this, but I must be doing something wrong
...
- 21
1
vote
1
answer
40
views
How does batch normalization make a model less sensitive to hyperparameter tuning?
Question 22 of 100+ Data Science Interview Questions and Answers for 2022 asks What is the benefit of batch normalization?
The first bullet of the answers to this is The model is less sensitive to ...
- 172
0
votes
1
answer
69
views
Why doesn't batch normalization 'zero out' a batch of size one?
I'm using Tensorflow. Consider the example below:
...
- 1
0
votes
1
answer
356
views
Equations in "Batch normalization: theory and how to use it with Tensorflow"
I read the article Batch normalization: theory and how to use it with Tensorflow by
Federico Peccia.
The batch normalized activation is
$$ \bar x_i = \frac{x_i - \mu_B}{\sqrt{\sigma_B^2 + \epsilon}} $$...
- 123
1
vote
1
answer
251
views
Explanation of Karpathy tweet about common mistakes. #5: "you didn't use bias=False for your Linear/Conv2d layer when using BatchNorm"
I recently found this twitter thread from Andrej Karpathy. In it he states a few common mistakes during the development of a neural network.
you didn't try to overfit a single batch first.
you forgot ...
- 113
1
vote
1
answer
2k
views
To freeze or not, batch normalisation in ResNet when transfer learning
I'm using a ResNet50 model pretrained on ImageNet, to do transfer learning, fitting an image classification task.
The easy way of doing this is simply freezing the conv layers (or really all layers ...
0
votes
1
answer
65
views
Batch normalization for multiple datasets?
I am working on a task of generating synthetic data to help the training of my model. This means that the training is performed on synthetic + real data, and tested on real data.
I was told that batch ...
- 111
1
vote
1
answer
77
views
Using batchnorm and dropout simultaneously?
I am a bit confused about the relation between terms "Dropout" and "BatchNorm". As I understand,
Dropout is regularization technique, which is using only during training.
...
- 23
3
votes
1
answer
5k
views
How batch normalization layer resolve the vanishing gradient problem?
According to this article:
https://towardsdatascience.com/the-vanishing-gradient-problem-69bf08b15484
The vanishing gradient problem occurs when using the sigmoid ...
- 363
1
vote
1
answer
206
views
Poor CNN performance after implementing BatchNormalization
I am training a CNN to classify malware images from a dataset named Malimg. Before implementing the BatchNormalization layer, I was getting an accuracy of 95.57% (see below for the graph of loss/...
- 155
1
vote
1
answer
213
views
Can Batch Normalization replace tanh in RNN?
Question
Can Batch Normalization (BN) be inserted in RNN after $x_t@W_{xh}$, and after $h_{t-1}@W_{hh}$ to remove $f=tanh$ and bias $b_h$? If possible, will this ...
- 585
1
vote
0
answers
64
views
BUG: Tensorflow accuracy on training data different when training than when evaluating
I think I have made some sort of error because my model has a substantially different accuracy while training:
...
- 111
1
vote
0
answers
322
views
How to properly train batch-normalization networks with gradient accumulation?
Using gradient accumulation efficiently replicate training with larger batch sizes for networks that are independent on the batch size. However, networks with BN layers rely on a batch size -- running ...
3
votes
1
answer
897
views
How does batch normalization work for convolutional neural networks
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, ...
- 109
2
votes
1
answer
2k
views
Batch normalization for image CNN - Why not use the mean of the entire batch?
Question
For CNN to recognize images, why not use the entire batch data, instead of per feature, to calculate the mean in the Batch Normalization?
When each feature is independent, need to use per ...
- 585
5
votes
1
answer
2k
views
Transfer Learning for CNNs and Batch Norm Layers
In some transfer learning models, we set the training argument to False to maintain the pre-trained values of Batch Normalization for example, but the ...
- 243
1
vote
1
answer
404
views
Why does batchnorm1d in Pytorch compute 0 with the following example (2 lines of code)?
Here is the code
...
- 111
1
vote
0
answers
124
views
Normalizing historical data in time-series LSTMs
I am currently trying to solve a sequence prediction problem using LSTMs in a keras architecture. To illustrate the problem I give the following example which resemble the problem I must solve.
Lets ...
- 131
1
vote
0
answers
273
views
How to standardise/normalise csv data in a tensorflow 2 dataset
Does anyone know how to apply standardisation (or normalisation) on the numerical features of a dataset loaded in batch via tf.data.experimental.make_csv_dataset in ...
- 111
3
votes
1
answer
2k
views
Keras BatchNormalization axis
I use spectrogram as input to a Convolutional Neural Network I have created with tensorflow.keras in Python.
Its shape is (time, frequency, 1).
The input's shape of the CNN is (None, time, ...
1
vote
1
answer
844
views
Which batch size to use when Batch Normalization?
I want to train a CNN in Keras (optimizer Adam) and by using batch normalization after every ConvLayer and before every activation layer.
So far I mostly see examples in which training is carried out ...
- 393
2
votes
3
answers
718
views
Normalization in production
I am currently writing a machine learning pipeline for my time series application. At the end of each month, I get the data gathered, normalize it ([0, 1]), retrain the ML model with the new ...
- 21
1
vote
1
answer
2k
views
Why would batch normalization allows us to use higher learning rate in the neural network?
I am doing some study about the BatchNormalization:
https://towardsdatascience.com/batch-normalization-8a2e585775c9
In the article, it says:
...
- 2,735
0
votes
1
answer
5k
views
The added layer must be an instance of class Layer. Found: <keras.layers.normalization.BatchNormalization object at 0x0000024B0C16B780>
I don't know why BatchNormalization is giving the following error. Googled out but couldn't find any relevant answers
My code appears to be this:
- 16
1
vote
0
answers
137
views
BatchNorm vs InstNorm from the perspective of feature distributions
What I understand so far...
The main purpose of BatchNorm is to overcome covariance shift -- more specifically what the authors of BatchNorm coined "internal covariance shift".
Covariance shift is ...
- 111
1
vote
1
answer
864
views
Batch Normalization with CUDNN
I want to introduce Batch Normalization in my C++/CUDNN implementation of CNN. The implementation is currently performing well (without BN) on the MNIST dataset.
I am using the CUDNN implementation ...
1
vote
0
answers
80
views
L1-Norm Batch Normalization for Efficient Training of Deep Neural Networks
Could anyone help to derive equation (15) dL/dx in L1-Norm Batch Normalization for Efficient Training of Deep Neural Networks ?
I found that the term inside the rectangle for the expression (12) is ...
- 283
1
vote
1
answer
708
views
Batch Normalization and Dropout together causing incorrect segmentation results
So, I've been running a test to see how well a number of networks can perform road segmentation on a particular customer's dataset. I am testing UNET, RDRCNN, and Tiramisu against each other.
UNET ...
- 31
2
votes
1
answer
155
views
Where can Batch Normalization be used? CNNs or everywhere?
Should BatchNormalization be used only in CNNs or can they be used in Fully Connected Networks, Recurrent networks as well?
- 724
1
vote
1
answer
490
views
How does Batch Normalization in Machine Learning address covariate shift and speed up training?
In this video and this answer, it's mentioned that batch normalization doesn't allow the mean and variance of the parameters of any particular hidden layer to vary too much with change in previous ...
- 299
1
vote
0
answers
164
views
Why does BatchNorm on FC layers increase my error?
I am training a deep CNN for multivariate regression, with three fully connected layers on top of the convolutions. I am using Sigmoid activation for FC layers. When adding BatchNormalization (BN) I ...
- 111
2
votes
2
answers
2k
views
Keras models break when I add batch normalization
I'm creating the model for a DDPG agent (keras-rl version) but i'm having some trouble with errors whenever I try adding in batch normalization in the first of two networks.
Here is the creation ...
- 23
0
votes
2
answers
4k
views
When to use model.train() vs model.eval() in Pytoch?
I have a model that is used in a reinforcement learning algorithm for checkers, a la AlphaZero. Similar to that network, mine features batch normalization after each convolution layer. I am aware that ...
- 147
0
votes
1
answer
1k
views
What exactly is BatchNormalization() in Keras?
A month or two straight away into building image classifiers, I just sandwiched the BatchNormalization layer between conv2d. I wonder what it does, but I have seen my model learn faster in presence of ...
- 319
0
votes
1
answer
336
views
0
votes
0
answers
201
views
Neural Network fails to train when a particular batch normalization layer is removed?
Background info:
I built a baseline CNN model for the cifar10 dataset using batch normalization and then ReLu activation after each convolution layer. There are a couple of max pooling layers in ...
- 11
1
vote
1
answer
2k
views
If we are using batch normalization as the first layer, can we forego standard scaling of inputs?
It is common practice to use the standard scaler on the inputs before feeding it to a deep learning architecture. I was wondering whether it is necessary if the first layer is a batch normalization ...
- 131
3
votes
1
answer
2k
views
Keras multi-gpu batch normalization
1) How does the batch normalization layer work with multi_gpu_model?
Is it calculated separately on each GPU, or is somehow synchronized between GPUs?
2) Which batch normalization parameters are ...
- 1,039
1
vote
0
answers
313
views
Do batch norm makes sense for regression problems?
When my network is performing regression (like DQN) it makes sense to use batchnorm in network when output of my network should vary from [0, 100000]?
one way to tackle it is to normalize output but ...
- 295
1
vote
0
answers
38
views
How much batch effect is too much batch effect?
Algorithms such as ComBat/SVA are powerful tools for the removal of batch effects. Small batch effects can be confidently removed by these methods. But surely there must exist batch effects which are ...
- 61
1
vote
2
answers
2k
views
Conv bias or not with Instance Normalization?
It is well known that Conv layers that are followed by BatchNorm ones should not have bias due to BatchNorm having a bias term.
Using InstanceNorm however, the statistics are instance-specific rather ...
- 131
0
votes
2
answers
1k
views
Why BatchNormalization fails in Keras
I try to test ResNet approach on cifar10 dataset with the following python code:
...
- 217
11
votes
2
answers
14k
views
Batch normalization vs batch size
I have noticed that my performance of VGG 16 network gets better if I increase the batch size from $64$ to $256$. I have also observed that, using batch size $64$, ...
- 570
11
votes
2
answers
4k
views
Does batch normalization mean that sigmoids work better than ReLUs?
Batch normalization and ReLUs are both solutions to the vanishing gradient problem. If we're using batch normalization, should we then use sigmoids? Or are there features of ReLUs that make them ...
- 499