Like and thank you very much for these tutorials!!! I have just finished this exercise and my neural network works perfectly... Thamk you! First of these three sections over.
Let's explain in previous videos that the whole idea behind deep learning is the back propagation algorithm. So in this video where the viewers are expecting an explanation of the back propagation algorithm, let's write a bunch of lines and tell the viewers "I think you understood that this is the back propagation algorithm". Bravo.
Sir. You Are Amazing. Thank you very much. I know the NN theoretically but recently I needed to understand the Deep Learning process in few words.You've made my day. One Note: could you test the algorithm with not a perfect input and show the error.
Goodmorning sir, Why you used exactly 20 nodes for hidden layers? I did research on books but i hzvennt understand how to choose the correct number of nodes for hidden layers. Kind regards Doctor.
For W1=2*rand(20, 25)-1 i think it should be W1=2*rand(25,20)-1. According to your source code the first parameter i.e. 25 represent the number of the previous layer, whereas the second parameter i.e, represents the number of nodes related to the next layer.
It is correct since it is exactly the other way around: the first number represents the number of the next layer, whereas the second parameter represents the number of the previous layer. Also I believe that if you would switch these two parameters by mistake you would run into an error because of wrong matrix dimensions.
Thank you for the video. I have a question. What if your inputs and output to the network are sensors data, how can it be done because i am having some troubles
Hi , I understand the DL concept but have a question that how to choose hidden layers and hidden neurons. Is there any specific rule or trial and error?
hello, i have a question i have implemented the neccessary argumets to create XOR funcion with the code i have learned just by now. however when i apply [0;0] as it descibes the very first rule of XOR gate, i am getting 0.2500 in each segments while it must give me apprximately 1 in the first input. why i am facing that problem? edit: well i might saw the problem. just because all the inputs are 0 at the first rule of gates, none of this gate implementations will fit to your code due to the hidden layer values will be zero(for all output it is getting the equal sign which is 0.25). so i need to think about something else to implement gate configuration of ANN.
Hi, Nuruzzaman. Your video "Neural Network using Matlab" doesn't show its comments (at least in my PC). For this reason, I want to congratulate you for that great video. Your explanations are very clear. I wanted to create a more general version of your SGD_method() to solve a exercise that I must to do, however, when I arrived to the calculation of delta and deltaweight, I realized that both calculations would produce incompatible multiplication of matrices. Please let me show you: delta = output * (1 - output) * error; If I treat output like a array, I don't know the procedure to calculate delta. In my opinion it's impossible, because if (for instance) output has a order 2x1, (1 - output) also would have this order. Then we will have a multiplication of 2x1 and 2x1... and then we have the array error. Something similar happens with deltaWeight. I would like to know if you can make a video to show how to handle the training and testing of a perceptron with more than one output. My bigger doubt is: how is affected the matrix of weights when a error appears in (for instance) the output 1? The error can affect every element of the matrix or only the weights related with that output? Best regards and congratulations again!
Hello Again, I ran the example and updated it for a larger input set. The updated version works but the softmax function returns Nan after a few samples during training which affects the whole training session. I searched for a safe version of softmax but couldn't find one. I found one example with phyton np. I install phyton and implemented the function but getting errors such as "Conversion of MATLAB 'sym' to Python is not supported". I would be appreciated if you can suggest a safe softmax function implementation?
Hi, the video was very good. To set up the network, my input data is 1000 x 3 and my target data is 1000 x 1. The output of training and testing their values is more nan. Why is this happening? What are the reasons for this?
Hello Nuruzzaman Sir, can you explain how you arrived at error of hidden layers as the product of next weight and delta of next node? Please share the source if you are short of time.
This video is very helpful. I am doing a cryptanalysis on my research right now using deep neural network. May I ask if I will use this code on your MATLAB demo, on which part should I define my 128 input nodes and 128 output nodes? Yours has 25 input and 5 output nodes? Thanks for the help.
Your contents are well arranged. I have couple of questions. 1) what is the logic of exponential and division in softmax function? 2) In my case I have an image of 20x30 with output male/female, how many input and output nodes, I need?
Thank you. You asked a good question. 1) please plot the output of 'softmax' function separately. It will answer your question. It is difficult it express it properly using text. 2) for every pixel, there will be one input nodes. For 20*30 image, you need 600 input nodes. 3) There are two classes - (i) male and (ii) female. So two output nodes.
@@NuruzzamanFaruquis Thank you for your reply. Now I get it. Is it possible, you suggest some book/material where I can get a detailed and more complex example. I know this question is not related to this video but if you can help, I will appreciate. Thank you.
sir can i use this same code.. instead of our digit (matrix) can we convert an RGB image to binary - then with binary matrix suppose the already known output i will give which is same as in this video... If i do this then how much epoch will i need if i will give 20 images of 28*28 pixels..?? please answer sir
Teacher, there is something I don´t understand. In the section of adjustment at the DeepLearning function: adjustment_of_w4 = alpha*delta*output_of_hidden_layer3'; adjustment_of_w3 = alpha*delta3*output_of_hidden_layer2'; adjustment_of_w2 = alpha*delta2*output_of_hidden_layer1'; adjustment_of_w1 = alpha*delta1*reshaped_input_Image'; alpha is a real number, delta is a (5x5) matrix multiplied for (output_of_hidden_layer3') wich is a (1x20) matrix... I don´t get this part And similar for the rest: delta 3 is a (20x5) matrix, and (output_of_hidden_layer2') is a (1x20) matrix delta 2 is a (20x5) matrix, and (output_of_hidden_layer1') is a (1x20) matrix delta1 is a (20x5) matrix, and (reshaped_input_Image';) is a (1x25) matrix How are this multiplication among matrix possible? and I know it works perfectly because I did the exercise Hope you might read and answer this message. Thanks in advance! Atte, Merlis.-
Thank you for you nice presentation. However, during running the code with Matlab2016a I found error as follow function [w1, w2, w3, w4] = DeepLearning(w1, w2, w3, w4, input_Image, correct_Output) ↑ Error: Function definitions are not permitted in this context.
function y = ReLU(x) ↑ Error: Function definitions are not permitted in this context.
Undefined function or variable 'ReLU'. >> TrainingNetwork Undefined function or variable 'ReLU'. Error in DeepLearning (line 9) output_of_hidden_layer1 = ReLU(input_of_hidden_layer1); Error in TrainingNetwork (line 46) [w1, w2, w3, w4] = DeepLearning(w1, w2, w3, w4, input_Image, correct_Output); Could you suggest where is the problem! Thanks in advance.
I am chemical engineer, I am new to this machine learning programs. My thesis involves ANN. I have the same question as natalia chu has. How do I decide no. of neurons in hidden layer?
It's me from the other video, Thank you soo much for you tutorials and explanations, I think I'm starting to understand how I should do it!! one other thing... in the case I'm working I have 2000 input nodes and only a binary output... 1 or 0 how should the correct output be?! only one value?! I don't know I can I make it end with one single node!
Thank you so much for your sharings :) 👏👏 but I have a question, if I use correct_output(:,:,k)', is it true or if I use input_image(k,:)', is it possible? I couldn't understand the matrices what you mean
Hello, I have one question. Is your delta corresponding to the generalized delta rule? I can’t find the derivative of ReLU function Good tutorial!! It’s very helpful for me👍 Thank you very much for sharing
Thank you so much for your lecture! Do you have any example on how to do semantic segmentation of an object in an image? And what if the input image is not a grayscale or binary? For example RGB or RGBD image?
Right now I do not have any example on semantic segmentation. However, I will try to prepare one for you. And and RGB image has three different values for single pixel. The pixel values are the input signal. We cannot take three input signal at the same time in a single input node. That's why we use grayscale image.
Hi! I truly enjoy your video. I even tried to run the same code in MathLab R2016a, but it is given me an error message "Not enough input arguments reshaped_input_image = reshaped(input_image(:, :, k), 25, 1); " I want to use this method for face recognition, because I am conducting research on a topic Improving the effectiveness of Deep Learning Approach for Face Recognition . How can I see the designed Architecture of this Algorithm in the Matlab that is the Network Architectural design and how to input face images into the network?
Thank you for the well prepared video. But this is just the conventional backpropagation algorithm known since 30 years. I did the same using basic language more 23 years ago. It would be better if you show how to use deep learning ready functions in Matlab. No one need to write the details of coding nowadays. Thank you
I agree with you. With such experience you do not need the details of coding. However, I have some enthusiastic new learners around. They want to learn from the very basics. Thank you for your valuable suggestion.
@@NuruzzamanFaruquis I have a question, why do the weights become NAN if we set 1 to more than one value in the outputs? for example if the first output is [0 1 1 0 0], the weights become NAN. Could you please explain it?
thank you so much sir, i tried exactly the sam ecode but it says that "Error using - Matrix dimensions must agree." can you help me with this problem? thank you in advance
ataScience New York 2019(January 23-24) Join us in New York, January 23-24, for this unique peer-to-peer event for top data science minds. Explore how to drive innovation within your business through a data-driven culture, build a world-class team, and utilize the latest innovation to solve some of the toughest problems facing corporate America. Book your tickets: bit.ly/2OWuepR
any chances to get the codes and the files from the video ? btw, we dont test our model with the same data we trained, it does not make sense, it is totally biased. you should have been created another data set for testing. and what about the validation of the model you have been trained. so here u have just only train the model with just 5 input without any validaiton and then use the same 5 inputs as showing testing which is not at all neither can be also as validation.
I think one can take any number of nodes from structural point of view which will only effect the number of weights. However, this selection will impact the performance of model.
thank you so much can you give me solution of this exo with matlab plaes Carry out the training of multilayer networks allowing to estimate the models in the following cases: 1. Function approximation: y ̂ = 𝑓 (𝑦) using the dataapproxi.mat data file 2. Prediction problem: y ̂ (t) = 𝑓 (y(t - 1), y(t - 2),…, y (t- 𝑛)) using the data file: dataestima.mat 3. Problem of identification of a dynamic model y ̂ (t) = 𝑓 (u(t - 1),…, u(t - 𝑛u), y (t - 1),…, 𝑧y(t - 𝑛y)) using the file of data: dataident.mat
@@rise4above You can email me at faruquizaman27@gmail.com. Please add 'Neural Network' in the subject, and include your identity (name, designation) at the end of the email.
You are one of the greatest explainer of deep learning, thank you very much, gazakallahokhiran, from egypt
Like and thank you very much for these tutorials!!!
I have just finished this exercise and my neural network works perfectly... Thamk you!
First of these three sections over.
Let's explain in previous videos that the whole idea behind deep learning is the back propagation algorithm. So in this video where the viewers are expecting an explanation of the back propagation algorithm, let's write a bunch of lines and tell the viewers "I think you understood that this is the back propagation algorithm". Bravo.
Great. you are a high level lecturer. I have got more benefits from this video. Many thanks.
You are truly a genius, May God increase you in knowledge
Thank you! You saved my life and project! Big love!!!!!
Congratulations to your really well made videos
thank you for these tutorials! the visuals are super helpful, there no way i could understand this without them you have an awesome teching method!!!
Excellent work thanks
A solid guide to deep learning
Thank you
@@NuruzzamanFaruquis Why 20 nodes in the hidden layer?
thanks for sharing your valuable lectures
Bests of the best. Great video
Thank you Erum Ashraf.
permission to learn sir. thanks you. from Indonesia
Sir. You Are Amazing. Thank you very much. I know the NN theoretically but recently I needed to understand the Deep Learning process in few words.You've made my day.
One Note: could you test the algorithm with not a perfect input and show the error.
this is a great point to start . thank you so much ,
You are welcome. Feel free to ask if you face any problem.
Hi! Could you put a video about preparing a datastore for a multiple-input image classification CNN in Matlab? Thanks
Thank you very much it's very helpful for me.
Knowledgable....& informative topics..... carry on.
Thank you
indeed a good tutorial to have an idea about deep learning
Thank you
very informative. looking forward for more videos.
Thank you. I hope this lesson will help you to understand the concept of DNN and train DNN easily.
Thank you for these tutorials!
Goodmorning sir,
Why you used exactly 20 nodes for hidden layers? I did research on books but i hzvennt understand how to choose the correct number of nodes for hidden layers.
Kind regards Doctor.
For W1=2*rand(20, 25)-1 i think it should be W1=2*rand(25,20)-1. According to your source code the first parameter i.e. 25 represent the number of the previous layer, whereas the second parameter i.e, represents the number of nodes related to the next layer.
It is correct since it is exactly the other way around: the first number represents the number of the next layer, whereas the second parameter represents the number of the previous layer. Also I believe that if you would switch these two parameters by mistake you would run into an error because of wrong matrix dimensions.
Thanks. It's a really great tutorial
Plz take lecture on hopfield neural network to train the image
I will try.
Can I give any 2d cad geometry data as an input to deeplearning in matlab??
Yes, you can. You have to process your data accordingly.
@@NuruzzamanFaruquis can you give some ideas that how to start it?? Should I convert all 2d cad geometry into image file or what???
Why are they 20 nodes in the hidden layer?
Please answer
25-5=20
Great Presentation,
thanks for the videos. this is useful during lockdown haha
Glad you liked it!
@@NuruzzamanFaruquis Why are they 20 nodes in the hidden layer?
Please answer
Informative
Carry on genius
Waiting for ur nxt Video
Thank you
Hi, thank you for this video. can you send me the datasets your used for this application
Thank you very much , but i have quastion when i write save ('neoral network.mat') and load('neoral network.mat')
Matlap saing this error !why??
Thank you vert much, well done and very useful! I hope you will insert other videos about Deep Learning and Machine Learning using Matlab.
Thank you for the video. I have a question. What if your inputs and output to the network are sensors data, how can it be done because i am having some troubles
You need a data processing module in between the sensor node and the input nodes of the network. That's all you need to do.
@@NuruzzamanFaruquis Data processing module?
@@NuruzzamanFaruquis I have 3 inputs data (1 reference and 2 sensor data) and 1 output data
I wrote the code like you did in the video but i had error, how can i solve it??
thank you its very good idea
Hi , I understand the DL concept but have a question that how to choose hidden layers and hidden neurons. Is there any specific rule or trial and error?
hello,
i have a question i have implemented the neccessary argumets to create XOR funcion with the code i have learned just by now. however when i apply [0;0] as it descibes the very first rule of XOR gate, i am getting 0.2500 in each segments while it must give me apprximately 1 in the first input. why i am facing that problem? edit: well i might saw the problem. just because all the inputs are 0 at the first rule of gates, none of this gate implementations will fit to your code due to the hidden layer values will be zero(for all output it is getting the equal sign which is 0.25). so i need to think about something else to implement gate configuration of ANN.
Very helpful, thank you!
May the elder gods bless you with even more knowledge, thank you thnk you thnk you
Hi, Nuruzzaman. Your video "Neural Network using Matlab" doesn't show its comments (at least in my PC). For this reason, I want to congratulate you for that great video. Your explanations are very clear.
I wanted to create a more general version of your SGD_method() to solve a exercise that I must to do, however, when I arrived to the calculation of delta and deltaweight, I realized that both calculations would produce incompatible multiplication of matrices. Please let me show you:
delta = output * (1 - output) * error;
If I treat output like a array, I don't know the procedure to calculate delta. In my opinion it's impossible, because if (for instance) output has a order 2x1, (1 - output) also would have this order. Then we will have a multiplication of 2x1 and 2x1... and then we have the array error.
Something similar happens with deltaWeight.
I would like to know if you can make a video to show how to handle the training and testing of a perceptron with more than one output. My bigger doubt is: how is affected the matrix of weights when a error appears in (for instance) the output 1? The error can affect every element of the matrix or only the weights related with that output?
Best regards and congratulations again!
Hello Again, I ran the example and updated it for a larger input set. The updated version works but the softmax function returns Nan after a few samples during training which affects the whole training session. I searched for a safe version of softmax but couldn't find one. I found one example with phyton np. I install phyton and implemented the function but getting errors such as "Conversion of MATLAB 'sym' to Python is not supported". I would be appreciated if you can suggest a safe softmax function implementation?
Hi, the video was very good. To set up the network, my input data is 1000 x 3 and my target data is 1000 x 1. The output of training and testing their values is more nan. Why is this happening? What are the reasons for this?
can you make a video explaining all the basic things like nnet, epoch, newp etc and code of training neural network in matlab
Sir, can i use this method for estimation of heat for my batch reactor temperature control ? Please respond
Did you use sigmiod to calculate the weight as taught in the clip first?
Really it's amazing...
How can I work out which input is the most effective / had the largest meaning to the output of a neural network
Hello Nuruzzaman Sir, can you explain how you arrived at error of hidden layers as the product of next weight and delta of next node?
Please share the source if you are short of time.
Sir, how to recognition images in video scene using deep learning in matlab
This video is very helpful. I am doing a cryptanalysis on my research right now using deep neural network. May I ask if I will use this code on your MATLAB demo, on which part should I define my 128 input nodes and 128 output nodes? Yours has 25 input and 5 output nodes? Thanks for the help.
Your contents are well arranged. I have couple of questions. 1) what is the logic of exponential and division in softmax function? 2) In my case I have an image of 20x30 with output male/female, how many input and output nodes, I need?
Thank you. You asked a good question.
1) please plot the output of 'softmax' function separately. It will answer your question. It is difficult it express it properly using text.
2) for every pixel, there will be one input nodes. For 20*30 image, you need 600 input nodes.
3) There are two classes - (i) male and (ii) female. So two output nodes.
@@NuruzzamanFaruquis Thank you for your reply. Now I get it. Is it possible, you suggest some book/material where I can get a detailed and more complex example. I know this question is not related to this video but if you can help, I will appreciate. Thank you.
Thanx How can I obtain equation model from nn-tool matalb
very helpful, thank u!!
You are welcome :)
sir can i use this same code.. instead of our digit (matrix) can we convert an RGB image to binary - then with binary matrix suppose the already known output i will give which is same as in this video... If i do this then how much epoch will i need if i will give 20 images of 28*28 pixels..?? please answer sir
Teacher, there is something I don´t understand.
In the section of adjustment at the DeepLearning function:
adjustment_of_w4 = alpha*delta*output_of_hidden_layer3';
adjustment_of_w3 = alpha*delta3*output_of_hidden_layer2';
adjustment_of_w2 = alpha*delta2*output_of_hidden_layer1';
adjustment_of_w1 = alpha*delta1*reshaped_input_Image';
alpha is a real number, delta is a (5x5) matrix multiplied for (output_of_hidden_layer3') wich is a (1x20) matrix... I don´t get this part
And similar for the rest:
delta 3 is a (20x5) matrix, and (output_of_hidden_layer2') is a (1x20) matrix
delta 2 is a (20x5) matrix, and (output_of_hidden_layer1') is a (1x20) matrix
delta1 is a (20x5) matrix, and (reshaped_input_Image';) is a (1x25) matrix
How are this multiplication among matrix possible? and I know it works perfectly because I did the exercise
Hope you might read and answer this message.
Thanks in advance!
Atte,
Merlis.-
Thank you for you nice presentation. However, during running the code with Matlab2016a I found error as follow
function [w1, w2, w3, w4] = DeepLearning(w1, w2, w3, w4, input_Image, correct_Output)
↑
Error: Function definitions are not permitted in this context.
function y = ReLU(x)
↑
Error: Function definitions are not permitted in this context.
Undefined function or variable 'ReLU'.
>> TrainingNetwork
Undefined function or variable 'ReLU'.
Error in DeepLearning (line 9)
output_of_hidden_layer1 = ReLU(input_of_hidden_layer1);
Error in TrainingNetwork (line 46)
[w1, w2, w3, w4] = DeepLearning(w1, w2, w3, w4, input_Image, correct_Output);
Could you suggest where is the problem! Thanks in advance.
I have rechecked the code. It is working properly in my end. Would you please switch to MATLAB 2018B or layer version and try again?
well organized video
teşekkür ederim kardeşim, anlaşılır şekilde anlatmışsın, eline sağlık
Kırıkkale Üniv. Bilgisayar Müh. Doktora Talebesi
Muhammed Ali ERBIR
i'm in cukurova university ,Turkey , ADANA are you working on image processing like object detection
Excellent.
Thank you
I have a numeric matrix as input file for training. Can I train that using deep learning
Yes of course. Why not?
I am chemical engineer, I am new to this machine learning programs. My thesis involves ANN. I have the same question as natalia chu has. How do I decide no. of neurons in hidden layer?
I am still studying about it. I will share my findings in upcoming lessons.
what if we feeed an input image which is not of "1" but close to "1" ?
hi , I need a help for preparing an RCNN multiclasse (multi object detection ) code in matlab can you help me please
Fine, your explanation. Can you tell in which ways (CNN or, Sparse Autoencoders or RBMs or LSTM are used to implement this Deep Learning?
Thank you sir can i do this code for gray image ? ???
It's me from the other video, Thank you soo much for you tutorials and explanations, I think I'm starting to understand how I should do it!! one other thing... in the case I'm working I have 2000 input nodes and only a binary output... 1 or 0 how should the correct output be?! only one value?! I don't know I can I make it end with one single node!
Take two output nodes - [1 0] for true and [0 1] for false. Otherwise even for no signal, there will be confusing false result.
Was helpful...One question: How to back propagate when there are also biases with the weights?
thank you very much Ifound it as best lecture would you please tell me how to realize a communication system on deep learning?
sir..can you provide code for image classification using DNN
Sure I can. Would you please tell me which classification (binary or multi-class) you are talking about?
@@NuruzzamanFaruquis multi class mr image classification i need
Hello sir .. iam Appling phd .. I need your help please
Can you explain neural networks algorithm with machine learning in prediction of anything
Are you asking about predictive model using neural network?
@@NuruzzamanFaruquis yes. In Matlab plz with code also. I'm beginner in learning all these concepts.
@@LoYal777Games I understand. I will try my best to prepare a tutorial that will help you building solid understanding.
Thank You for lecture! I am still puzzled - why exactly 20 neurons in hidden layer? Could You, please, provide formula for calculation?
Bro,y he is using only 20 nodes in hidden layer.Y is it so??
@@rajasekarsuriya6269 Hi, did you got the answer for how to optimize the number of neurons.
Thank you so much for your sharings :) 👏👏 but I have a question, if I use correct_output(:,:,k)', is it true or if I use input_image(k,:)', is it possible? I couldn't understand the matrices what you mean
Hello,
I have one question. Is your delta corresponding to the generalized delta rule?
I can’t find the derivative of ReLU function
Good tutorial!! It’s very helpful for me👍
Thank you very much for sharing
Derivative is 1 in non negative region. Elsewhere zero. Not defined at zero though.
thank you good work
can u upload for simple inputs and output values with normalization of variables?
can you please make a video on how to fix the optimum number of layer and what should be the neurons in these layers
Thank you so much for your lecture! Do you have any example on how to do semantic segmentation of an object in an image? And what if the input image is not a grayscale or binary? For example RGB or RGBD image?
Right now I do not have any example on semantic segmentation. However, I will try to prepare one for you.
And and RGB image has three different values for single pixel. The pixel values are the input signal. We cannot take three input signal at the same time in a single input node. That's why we use grayscale image.
@@NuruzzamanFaruquis Thanks a lot! So, for multi channel input we need to modify the DCNN structure...
@@NuruzzamanFaruquis sir can you please help me on a code to find wear rate for my project . i am really stuck???
The number of hidden neurons should be 2/3 the size of the input layer, plus the size of the output layer
Hi! I truly enjoy your video. I even tried to run the same code in MathLab R2016a, but it is given me an error message "Not enough input arguments reshaped_input_image = reshaped(input_image(:, :, k), 25, 1);
" I want to use this method for face recognition, because I am conducting research on a topic Improving the effectiveness of Deep Learning Approach for Face Recognition . How can I see the designed Architecture of this Algorithm in the Matlab that is the Network Architectural design and how to input face images into the network?
you dont have to run this file that shows error, run the the test file straight it will work.
run=deepnet then testdeepnet
Why you used exactly 20 nodes for hidden layers?
I have question please my laptop is cpu and i want to ran this code with data consists of arround around million samples ,can i do that
Yes. You can do it. However, the training process will take longer.
Thanks for nice tutorial. Please provide the ReLU.m file.
Thanks for the explanation ... Please, I want a code (neural networks) in order to check the quality of water ..... and God bless you good
have you written the code . i need help
Thanks for the useful instructions; I have a project which needs your help. How can I contact you?
Thank you for the well prepared video. But this is just the conventional backpropagation algorithm known since 30 years. I did the same using basic language more 23 years ago. It would be better if you show how to use deep learning ready functions in Matlab. No one need to write the details of coding nowadays. Thank you
I agree with you.
With such experience you do not need the details of coding. However, I have some enthusiastic new learners around. They want to learn from the very basics.
Thank you for your valuable suggestion.
@@NuruzzamanFaruquis I have a question, why do the weights become NAN if we set 1 to more than one value in the outputs? for example if the first output is [0 1 1 0 0], the weights become NAN. Could you please explain it?
thank you so much sir, i tried exactly the sam ecode but it says that "Error using - Matrix dimensions must agree." can you help me with this problem? thank you in advance
How can apply this for colored image ?! Instead of grayscaled image ?
This is a binary image. I have the same question. How can it be applied to grayscale image?
@@thelegend5248 I think by converting gray to binary and then by using binary matrix....
@@chetansoni8620 This idea is also fine. But actually I want to train deep learning with colored image. How to do that
why are there 20 hidden layers ?
ataScience New York 2019(January 23-24)
Join us in New York, January 23-24, for this unique peer-to-peer event for top data science minds. Explore how to drive innovation within your business through a data-driven culture, build a world-class team, and utilize the latest innovation to solve some of the toughest problems facing corporate America.
Book your tickets: bit.ly/2OWuepR
sir i also need code for image classification using DNN. can you provide it.
sir my email id is nitin1759@gmail.com
useful!
Thank you
how to implement lstm in matlab
I have plans to prepare contents on LSTM. Hopefully you will get your answer once the content will be ready.
sir please provide source of medical imaging
Please make a video on medical image i.e. .nii file
why 20 nodes??????????????????
any chances to get the codes and the files from the video ?
btw, we dont test our model with the same data we trained, it does not make sense, it is totally biased. you should have been created another data set for testing. and what about the validation of the model you have been trained. so here u have just only train the model with just 5 input without any validaiton and then use the same 5 inputs as showing testing which is not at all neither can be also as validation.
Thank you for this very helpful videos! Could you explain how you get the number of neurons in the hidden layers? Is there a specific rule?
Did anyone take the homework of "why 20 notes"?
Same question from my side? What is the answer?
I think one can take any number of nodes from structural point of view which will only effect the number of weights. However, this selection will impact the performance of model.
thank you so much
can you give me solution of this exo with matlab plaes
Carry out the training of multilayer networks allowing to estimate the models
in the following cases:
1. Function approximation:
y ̂ = 𝑓 (𝑦) using the dataapproxi.mat data file
2. Prediction problem:
y ̂ (t) = 𝑓 (y(t - 1), y(t - 2),…, y (t- 𝑛)) using the data file:
dataestima.mat
3. Problem of identification of a dynamic model
y ̂ (t) = 𝑓 (u(t - 1),…, u(t - 𝑛u), y (t - 1),…, 𝑧y(t - 𝑛y)) using the file of
data: dataident.mat
Yes. It is possible. However, I need more details to implement and train such a network
@@NuruzzamanFaruquis OK Mr
I have the all data fil.mat
@@NuruzzamanFaruquis can you send me your email address please
@@rise4above You can email me at faruquizaman27@gmail.com. Please add 'Neural Network' in the subject, and include your identity (name, designation) at the end of the email.