Batch Normalization | How does it work, how to implement it (with code)

💪 Want to become a deep learning buff? 👇 Watch my full course on deep learning on YouTube

Batch normalization is the secret weapon that solves the unstable gradients problem for many of the Deep Learning architectures. Let's take a look into how batch normalization works under the hood, what other benefits it has and how we can implement it using Keras on a Jupyter Notebook.

RESOURCES:

COURSES:

[misraturp]

🤔 Curious about deep learning?
Start with the Fundamentals of Deep Learning booklet to learn the essentials in 25 pages

[natarajanlalgudi]

Short, simple, clean and lucidly explained. Thanks so much.

[hsoley]

Your channel and the contents are amazing, keep posting new episodes. Best wishes from NYC

[Ekskwkwkwkw]

Mısra, Thanks for everything. Teşekkür ederim çokça :) I hope for you the happiest, healtiest life

[adityabillore]

Superb explanation, i have watched many videos but this is the best one 🔥🔥🔥

[FRANKWHITE]

Thanks for sharing! 👍 just subscribed 🙌

[Anteater]

How to decide the optimal number of hidden layers and neurons for a neural network? For each option do
I just train the network and then test it on test data and compare the results? Is this sufficient or do you have to do say 5-fold cross validation to get average test results for each choice of parameters?

[rajkumarsrini]

It's really simple and understandable, thank you

[SwarnaliMollickA]

How to use Batch Normalization before the input? Please reply. 6:21

[Aditi]

cool content and presented well..thankyou!!

[lukegloege]

Misra, this was a nice explanation! I have used batch norm after the input layer in lieu of normalizing the entire training dataset. Do you know if these two are always equivalent or if batch norm instead of normalizing the whole data is considered best practice? Please let me know if you know of any papers that discuss this. Thanks, I think you make create content, keep it up!

[lakeguy]

speaking about features / X values; how do you choose between scaling between 0 & 1 or standardizing with mean and variance? I enjoy your videos!

[virajvaitha]

Great explanation and example thank you

[hssp]

why did you keep two activation func, as a separate layer and also within dense?

[anilkr]

Hey, i am from commerce background if i learn those skils which require to be a Data analyst, will there be any problem for me getting job becouse of my Commerce background or from a Tech background???? Please help or make a video on this topic