Layers in a Neural Network explained

In this video, we explain the concept of layers in a neural network and show how to create and specify layers in code with Keras.
🕒🦎 VIDEO SECTIONS 🦎🕒
00:00 Welcome to DEEPLIZARD - Go to deeplizard.com for learning resources
00:30 Help deeplizard add video timestamps - See example in the description
05:46 Collective Intelligence and the DEEPLIZARD HIVEMIND
💥🦎 DEEPLIZARD COMMUNITY RESOURCES 🦎💥
👋 Hey, we're Chris and Mandy, the creators of deeplizard!
👉 Check out the website for more learning material:
🔗 deeplizard.com
💻 ENROLL TO GET DOWNLOAD ACCESS TO CODE FILES
🔗 deeplizard.com/resources
🧠 Support collective intelligence, join the deeplizard hivemind:
🔗 deeplizard.com/hivemind
🧠 Use code DEEPLIZARD at checkout to receive 15% off your first Neurohacker order
👉 Use your receipt from Neurohacker to get a discount on deeplizard courses
🔗 neurohacker.com/shop?rfsn=648...
👀 CHECK OUT OUR VLOG:
🔗 / deeplizardvlog
❤️🦎 Special thanks to the following polymaths of the deeplizard hivemind:
Tammy
Mano Prime
Ling Li
🚀 Boost collective intelligence by sharing this video on social media!
👀 Follow deeplizard:
Our vlog: / deeplizardvlog
Facebook: / deeplizard
Instagram: / deeplizard
Twitter: / deeplizard
Patreon: / deeplizard
KZbin: / deeplizard
🎓 Deep Learning with deeplizard:
Deep Learning Dictionary - deeplizard.com/course/ddcpailzrd
Deep Learning Fundamentals - deeplizard.com/course/dlcpailzrd
Learn TensorFlow - deeplizard.com/course/tfcpailzrd
Learn PyTorch - deeplizard.com/course/ptcpailzrd
Natural Language Processing - deeplizard.com/course/txtcpai...
Reinforcement Learning - deeplizard.com/course/rlcpailzrd
Generative Adversarial Networks - deeplizard.com/course/gacpailzrd
🎓 Other Courses:
DL Fundamentals Classic - deeplizard.com/learn/video/gZ...
Deep Learning Deployment - deeplizard.com/learn/video/SI...
Data Science - deeplizard.com/learn/video/d1...
Trading - deeplizard.com/learn/video/Zp...
🛒 Check out products deeplizard recommends on Amazon:
🔗 amazon.com/shop/deeplizard
🎵 deeplizard uses music by Kevin MacLeod
🔗 / @incompetech_kmac
❤️ Please use the knowledge gained from deeplizard content for good, not evil.

[@deeplizard]

Machine Learning / Deep Learning Tutorials for Programmers playlist: kzbin.info/aero/PLZbbT5o_s2xq7LwI2y8_QtvuXZedL6tQU Keras Machine Learning / Deep Learning Tutorial playlist: kzbin.info/aero/PLZbbT5o_s2xrwRnXk_yCPtnqqo4_u2YGL

[@explorenfts]

I like that you break out topics into short video sections. Thanks for doing that!

[@deeplizard]

Glad to hear! Thanks for letting me know!

[@ianmcmill]

I would like that you take breaks >5ms inbetween sentences. :)

[@ammarulhaq9790]

Not gonna lie the whole playlist is helping me a lot and probably the best explanation on Ml and AI on youtube

[@chaoukimachreki6422]

Why aren't you gonna lie?

[@richi1235]

@@chaoukimachreki6422 Because he is a virtuous gentleman.

[@chaoukimachreki6422]

@@richi1235 XD hey Ex be nice , that was a genuine question.

[@miladsayad2935]

I have searched and even paid to some courses, but this is just the best! Your explanations are wonderful! God bless you!!

[@AJ-dy2bt]

It worth noting that the information of this course is well structured and very clear. I can recommend this course to anyone new to machine learning. Thanks a lot for this course!

[@miguelcastillo1742]

Thanks! Someone on Reddit recommended me this channel to get into Deep Learning. Now I have a better understanding what the Neural Networks are.

[@sanazheydari8970]

Dear Mandy you are wonderful, you are a real teacher that I' have seen in my life your videos are really super informative you can not imagine how much your videos helped me God bless you you will be in my mind forever, thanks a lot

[@siddarthkotthur1558]

I really thank you for making this free and available to everyone!

[@abdulhaseeb7081]

Wonderful! I was searching alot about layers but i could'nt find, and here is what i was searching for thanks deeplizard!

[@marvhan888]

parroting other comments : you are amazing teacher, talented with how you explain any technical topic, THANK U SO MUCH

[@mmesoemmanuel4410]

I love you.. You are the only source that explained ANN easily to me

[@tolifeandlearning3919]

Wonderful, easy explanation. Love this channel.

[@tymothylim6550]

Thank you very much for this video! This is the first time I understood how Convolution is a kind of layer in an Artificial Neural Network!

[@abagofcashews3234]

This really helped me understand what I was missing. Thanks!

[@rajuthapa9005]

Very informative, practical tutorials ever found. Thank you

[@Plerion]

I found the answer to a question I had in a recent comment, so thank you for taking your time in replying even until today.

[@sanwalyousaf]

I was struggling with the concept of dense layer a lot. This video makes it very easy. Thanks a lot.

[@PunmasterSTP]

Yeah me too. I think this material can be rather...dense. 😎Sorry I just had to.

[@faezakamran3793]

You are amazing I must say. You not only share the best tutorials but also answer people's questions and that too with great detail. Keep up the good work!

[@deeplizard]

Thank you, Faeza!

[@Jeremy-bd2yx]

dude these are incredible. thank you for the direct and conceise explanations.

[@azhaanali1109]

Thank you !! This is exactly what I was looking for

[@artarriaga7206]

This is so concise. Thank you for putting this together

[@tannusharma2285]

Thank you so much...Blessed to find this channel

[@reinasama904]

This is wonderful! Thank you so much.

[@afreenshaik7984]

Thank you so much for these clear explained videos ❤️👍

[@BillyBarnyarns]

This is a brilliant tutorial!

[@pradeeppaladi8513]

Hey @deeplizard, Thank you so much for answering my question that even ChatGPT couldn't answer it clearly. I am actually a beginner in deep learning. Your explanation has answered my question. In fact, I have subscribed to your channel as well!

[@shabnampathan4515]

Very nice explaination.Thank u so much .

[@michaelkaufman5119]

Your videos are infinitely better than anyone else. I'm might as well just search your channel instead of KZbin for these topics.

[@deeplizard]

Search the website also :D deeplizard.com

[@GelsYT]

Thank you so much, your explanation is very simple :) thank youu

[@jisoo9285]

Thank you so much for the video!

[@stanstanstan]

Incredible, I'm learning so much

[@firuzjuraev5383]

It is sooo amazing tutorial. I love the way you explained. Thank you soo much

[@mikhailsladkomedov9037]

Thanks, the video is awesome as well as the whole playlist!

[@ibrahimzahir7061]

This is great. Love this series. Hope to see more ML n deep learning tutorials

[@deeplizard]

Thank you, ibrahim! Glad you're enjoying the series! We have many other ML and DL series available here on KZbin and on deeplizard.com. You can browse all available series on the deeplizard.com home page!

[@scottk5083]

Excellent content as usual man. love it

[@poffer31]

This was useful video to understand layers, Thanks!

[@entertainmentandit422]

Great tutorial very helpful

[@qusayhamad7243]

thank you a lot for your help

[@ryadboukhari4801]

Very helpful, thanks (y)

[@praburocking2777]

good work, now I understood how NN in keras works,thanks for ur help.hope u do more videos like this.

[@deeplizard]

Thank you, Prabu! The full series is here: deeplizard.com/learn/playlist/PLZbbT5o_s2xq7LwI2y8_QtvuXZedL6tQU Many other NN series are on the website as well :)

[@hugodasilvadasilva]

Hi, thanks for your videos. Is there any prediction for videos about the different type of layers (convolutional, dense, pooling, etc)?

[@vidumini23]

Thank you soo much.

[@willberger96]

Great video!

[@mariamal2189]

thanks a lot for this video

[@richarda1630]

awesome vid guys! :)

[@cahyaramadhan2971]

amazing, thanks

[@sashamuller9743]

great videos!!

[@coolcalmnormal]

Thank you

[@AkshayGhadi01]

You are simply awesome.

[@classictremonti7997]

It is clear that the image "NN.PNG" was loaded into the program and displayed, however is there a way to generate a neural network architecture diagram based upon the code that implements it?

[@ranarehan9185]

i was also search for the same thing on the internet but i think now there must be any possibility to draw the same graph of your code if there is any and any one know kindly let me know in the comment

[@animeshsrivastava5737]

This is best channel to learn from period

[@animeshsrivastava5737]

The* ffs cant type for shit

[@nawnitsen7771]

Best tutorial for ML

[@vidhyak2102]

Amazing and crystal clear explanation. You are a great teacher. Do you plan to take a session on NLP

[@deeplizard]

Thanks, Vidhya! NLP is on my list of potential topics for future videos!

[@vidhyak7096]

awesome. Looking forward

[@laxminarayanchoudhary939]

Thank you, awesome tutorial. But, I have a doubt about multi-class problem, is it possible to have a model having 11 unit(features) input layer and 5 unit(classes) output layer with no hidden layer.

[@soundsbyel1421]

Ty.

[@yamanyaseen9922]

Hey deep lizard, how do we know that there's a need of dense layer for our network? Example, for image enhancement through neural network do we require any dense layers?

[@jabrams3643]

hi! im trying to get this sorted out after so long of not being able to grok it lol. so for every node in the INPUT LAYER we put in a "feature" of a single training example, is this correct? so if we were to train in batches does that mean that each neuron in the input layer gets that same specific feature passed in but for several different examples? (passed in as some sort of array with many items in it). thank u for your time

[@deeplizard]

Yes, you've got it, j!

[@AlgoTribes]

@deeplizard..why did we used np.expand_dims and why we are plotting the img using index img[0]...i know it is a silly question to ask but i don't mind asking..

[@PunmasterSTP]

Layers in a neural network? More like "Lizard, you make this material just work!" Thanks for teaching us in such a truly effective way.

[@lightningblade9347]

Hi there, before asking my question I wanna thank you for your amazing and well explained videos, my question is: how do we specify the number of neurons in the hidden layer ? Do we choose whatever number of neurons we want or is the number of the latter in the hidden layer tied to something (which I don't know) ?

[@deeplizard]

You're welcome, Lightning Blade! Generally speaking, the more complex your data, the more layers you can expect to have in your model, and the more neurons in each layer you'll likely need. Also, you'll typically see that the number of neurons increases within each layer as the layers become deeper in the network. There is not a general rule of thumb that I'm aware of to follow for choosing how many neurons to include in each layer. It's more of mixing trial and error along with experience from what's worked in previous models.

[@lightningblade9347]

Thank you very much for the explanation, I'll make sure to watch all of your deep learning videos, I might have some other questions in the coming videos which I hope it won't bother you. Thank you again.

[@goldensleeves]

Great videos! Question: Is the output layer an ACTUAL output like "Cat" and "Dog"? Or is the output layer technically another hidden layer but it just happens to be the last hidden layer so it's the one that determines the classification decision? Thanks for clarification!!

[@deeplizard]

The output is a probability distribution among possible classes. The class with the highest probability is the one that we choose as the predicted label from the network. This point will be clarified further as you progress through the course.

[@goldensleeves]

@@deeplizard hmm ok I am looking forward to the rest of the course! I guess I am having trouble distinguishing the difference between an output layer and an output. Hopefully that will clear up. Thanks so much for the response. Liked and subscribed. All the best

[@ice7mayu]

Amazing video, very helpful for understanding these concepts. Just a little piece of advice, the font size should be larger :D.

[@deeplizard]

Thanks, Yu! In later videos, I zoom in on the code :) Also, you can check out the corresponding blogs as well to get a better view of the text: deeplizard.com/learn/video/FK77zZxaBoI

[@ice7mayu]

@@deeplizard I'm reading the blog now, really they're very helpful

[@deeplizard]

Perfect! Glad to hear that :)

[@roirtur2186]

Hi, I understand how layers work and everything, we give a number of inputs, do a few operations on layer to get a certain number of outputs in the end. But how do you know how many hidden neurons to put? I looked online and couldn't find any explanation other than "it depends on what 'rule' you're using". So I can put how many I want? ok but what is the best number to use then? I found another saying "well, you try with more or less neurons and you check by hand". Is there any better way to find out ? If you have an idea... :) Ty

[@SuperKafooo]

I'm so glad that I came across your videos. However, I have one question about the 'input_shape' parameter. So what would be my input shape if I have, say 60 RGB 128x128 images of cats and dogs. Also what will it be if it's just gray scale?

[@deeplizard]

If the images were colored images, then the input shape would be (128,128,3). If they were gray scale, then the input shape would be (128,128,1). Check out our video at the link below starting at 3:03 for an illustration of this. deeplizard.com/learn/video/gmBfb6LNnZs

[@Viralvlogvideos]

please clarify my doubts mam, if the assigned weights changes for optimization then it will not get confuse?? and my 2nd doubt is when ur making dense layers in coding frist u used relu and then u used softmax?? why so

[@tusharbarman1924]

Please make a videos on NLP like word embeddings, LSTM, GRU etc

[@yikenicolezhang]

In thefigure 2:52, there is a line between the hidden layer neuron 1 and output layer neural 2 missing. Thanks for explaining!

[@deeplizard]

Good eye, you're right thanks!

[@DURGESH-gy8hb]

will you please advise for any book for (neural network). your teaching is best teaching .i am benefited.

[@deeplizard]

Thank you, Durgesh! I recommend this one: neuralnetworksanddeeplearning.com/ Also, be sure to check out deeplizard.com. There, we have corresponding text-based blogs for many videos on our channel.

[@kwantom1996]

In the NN.PNG graph, why doesn't the first neuron at the hidden layer connect to the second neuron at the output layer? Is it a misdrawn or intentional? Thank you.

[@deeplizard]

This was not intentional, and I didn't catch it until after the upload. Good eye :)

[@yohannnedelec9305]

Probably a dumb question, but in your neural network image, aren't we supposed to have an edge between the top node in the hidden layer and the bottom node in the output layer ?

[@deeplizard]

Hey Yohann - Not a dumb question at all! You're right, there should be an edge there. I didn't notice that missing detail! In later videos of the playlist, we start using another network as our visual, so this missing edge won't persist for too long.

[@zenchiassassin283]

:0 I didn't notice

[@22kmhigh]

Thanks for the question, I had to browse ~50 comments to make sure that this missing connexion was actually a missing connexion

[@owaisaos1851]

its a one of best tutorial about AI on youtube if it is possible for you then please make tutorial on tensorflow

[@deeplizard]

Thank you, Owais! I have Tensorflow on my list to explore as a possible topic/playlist for the future. For now, I have a Keras playlist that focuses on the coding-side of neural networks in case you haven't seen it yet. kzbin.info/aero/PLZbbT5o_s2xrwRnXk_yCPtnqqo4_u2YGL

[@SameeraDissanayakaUOM]

hi deeplizard I'm following this tutorial series,i love this but i try to genarate same layers graph using scipy it says module 'scipy.ndimage' has no attribute 'imread' any solution

[@chandlergreff3113]

Hey! for some reason, the activation functions dont work for me. It states when hovering over 'Activation' : "unused import statement" and says that "relu" is a typo. What should I do? Thx in advanced !

[@ivanzapote8369]

i just subbed bc of the channel name & pfp

[@deeplizard]

🦎😎

[@louerleseigneur4532]

merci

[@mdyeasinarafath4450]

Great explanation mam! Suppose I have 3 features, each feature has 4 possible values. Meaning they are not only confined between either 0 or 1. Is that okay for Neural Nets?

[@deeplizard]

Thanks, Md.Yasin Arafat Yen! Yes, having labels that aren't 0s and 1s is fine. Most of the time though, you will standardize the labels. You can see more about this in this video on preprocessing data for training a neural network: kzbin.info/www/bejne/i5zdmaKretBpq7s And I also discuss another labeling technique called one-hot encoding in this video: kzbin.info/www/bejne/rJCXfIqjoNygqbc

[@mdyeasinarafath4450]

Thanks! I meant my features are not binary! Suppose I have 3 features and 1 Class: A B C Class 1 1 2 1 2 0 1 0 0 2 0 2 Here I don't need any encoding right?! Does ANN works with this type of data?

[@deeplizard]

Yes, that's completely fine!

[@siddhichaukade]

Could you please create a playlist for Mask RCNN?

[@FantaPopRockz]

Great

[@gajendrasonare]

So n in Dense(n) depends on how many things you are predicting? I tried to put n = 2 in last dense layer of cnn model but it gave me error. Why?

[@Viralvlogvideos]

so every time the weight assigned is changing?

[@James-eb7ph]

So would there also be weights from the hidden layer to the output?

[@deeplizard]

yes

[@notlikethat1819]

what does the hiddon layer exactly hide?

[@user-ql9vz9un3f]

For those who follows the video use plt.imread right now, cause ndimage.imread is deprecated

[@draganatosic9638]

Thanks for another super informative video. I have one question: how are the weights of the connections between input- and hidden layers being determined? Thank you for your time and shared knowledge :)

[@deeplizard]

Hey Dragana - The weights are first randomly initialized, and then they are optimized during the training process with Stochastic Gradient Descent (SGD). SGD makes use of backpropagation to find the optimized weights. Don't worry if you're not yet familiar with these concepts. The videos below, which come later in this playlist, explain these topics in detail. Weight initialization: kzbin.info/www/bejne/bpzVlWingLuqY7M Backpropagation (5 part series): 1. kzbin.info/www/bejne/jnaWnKWcaKiEotU 2. kzbin.info/www/bejne/aJ62qqaIrZJkmZI 3. kzbin.info/www/bejne/fWbFZZ2Id7CBrtk 4. kzbin.info/www/bejne/kKOYp5x3j6yhmqc 5. kzbin.info/www/bejne/rnTPfJKVeNaNpLM Show less

[@draganatosic9638]

deeplizard Thank you!!! :)

[@augustvctjuh8423]

The top node of the hidden layer is supposed to be connected to the bottom output node, right?

[@deeplizard]

That's right. I didn't notice it was missing until much later.

[@mage_ex8910]

Nicee

[@Wenyfile]

When you explained input_shape I got a little confused. Did you mean that input_shape=(3 ,) means that you have 3 input neurons or did you mean that each neuron in the hidden layer will receive 3 signals each? Or maybe that's the same thing?

[@deeplizard]

Hey Christoffer - 3 input neurons. For example, if each sample in our training set was an array containing a person's age, height, and weight, then the network would have 3 input neurons. Each neuron represents a single feature from the sample. One for age, one for height, and one for weight.

[@Wenyfile]

Thank you! As a follow up question if you had for example 20 images of numbers (0-19) in this type of network would you only assign 1 neuron because all would be classified as numbers? I'm sure you would use another more effective technique for that but just so I understand correctly or maybe it doesn't work at all for that. I'm very new to deep learning so sorry if the question is confusing.

[@deeplizard]

For images, each individual pixel is considered a feature. So, if you had a grayscale image that was 20x20 pixels in size, then in total you would have 20 x 20 = 400 total inputs to the network representing a single image. You might be interested in the videos/blogs below where we expand on this topic further when we cover convolutional neural networks (CNNs). deeplizard.com/learn/video/k6ZF1TSniYk deeplizard.com/learn/video/gmBfb6LNnZs If you need a basic explanation of CNNs before jumping into those, then check this one out first: deeplizard.com/learn/video/YRhxdVk_sIs

[@chawza8402]

My question is, My do you chose 5 neurons in the hidden layer? why not other numbers and does it matter?

[@khadidjamoghrani197]

where can i find the code please?

[@ritvikpalvankar1903]

How do I choose weights between 2 hidden layers?

[@kartikthakkar2272]

What does a 'shape of the data' mean?

[@gezahorvat553]

{ "question": "How do you pass in layers into a sequential keras model?", "choices": [ "as an array", "as a list of function arguments", "as an object literal", "it isn't possible" ], "answer": "as an array", "creator": "Hivemind", "creationDate": "2020-10-20T22:01:48.221Z" }

[@deeplizard]

Thanks, Géza! Just added your question to deeplizard.com/learn/video/FK77zZxaBoI :) Note that I modified it to have an answer "as a list" and updated the written portion of the blog to reflect this, as technically the data type of what the Sequential model expects is a list, not an array. Thank you!

[@zaidalattar2483]

Actually>>>> you are the best ..... that's it .

[@Quran_with_translation]

Great Video.,But very hard to read text in the code.

[@sandyjust]

node output = activation(weighted sum of inputs) => this also need the 'bias' term right? So the actually it would be node output = activation(weighted sum of inputs+bias)

[@deeplizard]

Yes, bias is covered in detail here: deeplizard.com/learn/video/HetFihsXSys

[@sayyedulislamhridoy4734]

How to measure size of nodes in Hidden layer for example: Dense(units=5 .... ?

[@festivalblogTV]

Shouldn't each neuron in the hidden layer be connected to each neuron in the output layer? If so, the first neuron of the hidden layer may be missing a connection to the second neuron of the output layer. (...and thanks for these didactically excellent videos!)

[@deeplizard]

You're right, I didn't notice the image was missing a connection during the recording 😅

[@tantriduong4612]

Thank you so much for the excellent course. But I get this error when trying: " AttributeError: module 'scipy.ndimage' has no attribute 'imread' " It makes no sense. Please help me fix this, also can I do this in Pycharm (which came up with the same error statement)?

[@deeplizard]

ndimage.imread() has been deprecated. Use matplotlib.pyplot.imread() instead.

[@tantriduong4612]

@@deeplizard that's good to know, it worked, thank you very much

[@tonihuhtiniemi1222]

When running on Jupyter: "AttributeError: module 'scipy.ndimage' has no attribute 'imread'" Im I missing some package or what? Thanks!

[@mohammedesoofally]

CV2 or PIL

[@HossamKorin]

Hi, I tried plotting the neural network using the matplotlib code you used but I seem to be getting an attribute error"module 'scipy.ndimage' has no attribute 'imread'

[@upenderreddy7992]

Yes, even i got the same error in google colab and in jupyter it says FileNotFoundError: [Errno 2] No such file or directory: 'NN.PNG'