transformer neural network simply explained

#transformer #neuralnetwork #nlp #machinelearning
Hello, in this video I share a simple step by step explanation on how Transformer Neural Network work.
Timestamps
0:00 - Intro
0:45 - Understanding attention technique
1:40 - Problem with sequence networks
1:57 - Motivation for Transformer networks
3:22 - Positional Encoding
4:32 - Vanilla Attention
5:22 - Self Attention
6:52 - Multi-Head Attention
7:50 - Residual Connection & Normalization
9:53 - Masked Multi-Head Attention

[@mahibbalde1433]

Amazing visualization on this topic

[@wp1300]

3:00 1 Input Embedding
3:40 2 Positional Encoding
4:08 3 Encoder Layer
4:30 3a. Multi-head Attention
7:51 3.a.ii Residual Addition, Layer Normalization & Pointwise Feed-Forward
9:28 4 Output Embedding & Positional Encoding
9:33 5 Decoder Layer
10:07 5.a. Masked Multi-head Attention

[@williamkariampuzha7951]

It finally cleared up what attention is for me! Thank you so much!!

[@eniolaa]

I'm so glad!

[@akrsrivastava]

Great explaination.

[@sarahsalt3689]

Thank you for sharing!

[@Amapramaadhy]

Loved your explanation. What an amazing mix of clear presentation and insightful visualization! Please keep it up.

[@eniolaa]

Thank you kindly!

[@d3v487]

You teach very intuitively ❤️... Please make more videos on Deep leaning concepts , people definitely like.. : )

[@eniolaa]

Thank you, I will

[@lmnefg121]

this is soooo helpful~! Just love it

[@eniolaa]

Glad it was helpful!

[@sreebvmcreation9388]

Nice explanation

[@TosinAlabi]

Well done eniola, totally enjoyed your video.!!

[@eniolaa]

Thank you so much!

[@RicardoSilva-lm4hb]

you won a subscribe!

[@maqboolurrahimkhan]

Perfect thank you so much

[@eniolaa]

You're welcome 😊

[@cedricmanouan2333]

😭 i’m subscribing directly 😌

[@Dhirajkumar-ls1ws]

Thanks

[@hodgeshaywood3764]

Hi Great explanation! Which software did you use for your animation and presentation?

[@anshulkumarprajapati8346]

Very nice explained ma'am can you please make more videos on ttn to get more clear

[@satyamgaba]

So crispy and precise.
When you're stacking multiple decoders, will we stack masked multi head attention too?

[@eniolaa]

yes we do, the masked multi head attention unit is a part of the decoder.

[@coreygates1162]

Hello! wonderful explanation and video! I had one question, when going from the encoder layer to the embedding layer, what is the input to the output embedding layer? is it still "I am a student"? How are the French words incorporated?

[@coreygates1162]

I guess like for the output embedding, would the look up table be the learned french translations?

[@marcosmartinez9241]

what is recived in the output embeding?

[@aguntuk10]

ok How does query ,key and values calculated?

[@rahulsbytes]

By the Query ,Key and Values matrices which is randomly assigned in intial state.

[@mehmetaliozer2403]

I couldn't get the idea of masked multi-heead attention, any comment i would be appreciated.