How To Implement Artificial Neural Network (ANN) in SIMULINK ? (Part-2) | Dr. J. A. Laghari

[@aaronswai3092]

Be blessed Dr. Laghairi. I, personally, I have learned a lot on the implementation of the ANN in MATLAB

[@DrJALAGHARI]

Thanks for the compliments. Stay blessed

[@do_not_reply4784]

sir , i always tell my students to watch your video , to make some project/ publication in their final year. 2 of my group published with the help of your videos. thankyou, waiting for more content from you

[@DrJALAGHARI]

Thank you so much for the compliments.

[@easyhack5859]

Thank you dr sahb. Pray from Allah almighty to give me success. Wonderful methodology. Best teacher. Stay blessed. Love you from sindh Pakistan. Your accent is sindhi. Love you sir.

[@DrJALAGHARI]

Thanks for the compliments. Stay Blessed.

[@ozenozcamgoz5454]

Dr. Allah sizden razı olsun, I mean Allah give you and your family good life in here and also in the other side inşAllah, very grateful for your courses

[@DrJALAGHARI]

Thank you so much for the compliments.

[@luisgonzalez2020]

Felicidades por su excelente trabajo, y el esfuerzo que hace para crear material de gran calidad. Bendiciones

[@DrJALAGHARI]

Thank you so much for the compliments.

[@MuhammadArshad-yf6vd]

Sir the way you presents is really amazing. Thank you for the efforts.

[@DrJALAGHARI]

Thank you so much for the compliments.

[@abdulazizalbarazi1346]

Assalamualaikum sir, thank you very much for your great work, this is very helpful! Just an important notice I wanted to share with you regarding the layer size assignment [3,5,1]. The first one '3' is actually the number of neurons of the first hidden layer, and not the input layer. '5' would be for the second hidden layer and finally '1' is for the output layer. The input layer does not hold an activation function, it only holds the input values of the network and no processing is done here. I believe mentioning this detail is very important to avoid misconception and confusion. Thank you again and have a wonderful day

[@DrJALAGHARI]

Thank you so much for your detailed explanation. Infact you are right. I agreed.

[@lampfall7915]

Really clear and comprehensive. Great job. I really appreciate

[@DrJALAGHARI]

Thanks for the compliments.

[@lampfall7915]

Could you explain us about ANN optimizations?

[@CarlosHenriqueDomiciano]

This video has an incredible wealth of detail. I learned strongly. Thank you very much.

[@DrJALAGHARI]

Thank you so much for the compliments.

[@Electrical.Engineer730]

Thank you so much for this dear sir I really appreciate and respect of your efforts for us

[@DrJALAGHARI]

Thank you so much for the compliments. I hope it will help you

[@charlesokoh3373]

Thank you Sir for saving our modern study life💯🙏🏽

[@DrJALAGHARI]

Most welcome. Do share this within your circle so that they may also get benefit from this video as well as from this channel.

[@hazemal-otum5306]

Thank very much Dr. This tutorial was very helpful. May ALLAH bless you

[@DrJALAGHARI]

Thank you so much for the compliments. Most welcome.

[@kashifbangsh4242]

Best explanation. Keep it up.

[@DrJALAGHARI]

Thanks for the compliments.

[@mr.n.praneetheee1035]

really foundational and boosting.. thank you sir

[@DrJALAGHARI]

Thanks for the compliments. Don't miss the other parts of ANN series. I hope you will also enjoy those parts too.

[@afnanghanem324]

thank you very much for these useful videos. I really appreciate your efforts ,may Allah bless you .

[@DrJALAGHARI]

Thanks for the compliments. Do share it within your circle too.

[@hantr1998]

Thank you very much! This lecture was so useful !

[@DrJALAGHARI]

Thanks for the compliments.

[@imranghilzai1]

Very wonderful video... thanx alot for sharing your knowledge ...

[@DrJALAGHARI]

Thanks for the compliments.

[@imranghilzai1]

@@DrJALAGHARI sir I am wornking on ann technique so can you share email or whatsapp number so I can ask few questions??

[@crchetann]

Thank you so much Dr. Laghari

[@DrJALAGHARI]

Thanks for the compliments.

[@KanchanJhaPEE]

Thank you a lot Sir, this lecture is really helpful.

[@DrJALAGHARI]

Thanks for the compliments.

[@dr.nafeesahamad8567]

Dr. Laghari, Superb

[@DrJALAGHARI]

thanks for the compliments.

[@dr.kranthikiran2230]

Excellent explanation, sir.

[@DrJALAGHARI]

Thanks for the compliments.

[@Dieu_Seul_Suffit__PJG]

That's the best..... Thank you🍉🍍

[@DrJALAGHARI]

Thanks for the compliments.

[@iminasha1210]

salam Dr, may i know how to correct this situation? Undefined function 'minmax' for input arguments of type 'double'. Error in untitled6 (line 13) net=newff(minmax(I),[3,5,1],{'logsig','tansig','purelin'},'trainlm');

[@DrJALAGHARI]

What is the name of the input data in the workspace. Change the name accordingly

[@fahadamyal620]

Thank you, its very helpful tutorial, god bless you

[@DrJALAGHARI]

Thanks for the compliments.

[@Kollysman]

Thank you so much for this amazing tutorial sir. I want to train my ANN with Mayfly Algorithm on MATLAB, how possible is that to do sir? I'll appreciate your guide Dr.

[@DrJALAGHARI]

Thanks for the compliments.

[@mohammadijajahmed619]

Sir, your videos are great source of information. Could you please explain how to implement ANN from earthquake records? ( Geotechnical engineering point of view). it would be helpful for my masters project. Thank you.

[@DrJALAGHARI]

Thanks for the comment. If you have the data then you may use this data with the ANN to predict the earthquake.

[@debashishmondal9601]

Hello Dr. Laghari, In the input layer Y= 2a+3b+5c, why you use 2, 3 and 5

[@DrJALAGHARI]

Thanks for the comment. This is just an arbitrary example to make you understand the example. You can use any values of your choice.

[@Piyush-qn2ks]

5:05 matlab didnt just generate 10 random numbers between 10 and 15. it has also generated numbers beyond the range. kindly give correct code

[@DrJALAGHARI]

Thanks for the comment. Basically, It will generate the random numbers between 10 and 25. The code is corrrect.

[@easyhack5859]

JazakaAllah kher

[@DrJALAGHARI]

Thanks for the compliments.

[@yentrapragadasampoorna7607]

For example 3 a=A*rand(x, y) +B the input minimum and max ranges are 10 and 15 but the values are beyond 15 .. Please explain

[@DrJALAGHARI]

Thanks for the comment. Kindly watch the video step by step for better understanding.

[@paidiravi7723]

Thankyou sir giving these important points

[@DrJALAGHARI]

Most welcome

[@asmaahassanein724]

thank you soo much may allah rewards you for that inshallah

[@DrJALAGHARI]

Ameen Thank you so much for the compliments.

[@muhammadashraf6645]

sir can we tune PID controller with ANN??just like we tune with PSO or GA etc??

[@DrJALAGHARI]

Thanks for the comment. Yes.

[@sravankumar1263]

thankyou very much for the video sir,it is a great opportunity for me to gain the knowledge from your videos. sir under faulty coditions and laod variations how the min and max values can be choosen??

[@DrJALAGHARI]

Thanks for the compliments. For fault Conditions, you have to simulate the power system for all types of fault and get that data. When all types of faults are simulated at every location then that data will include all the minimum and maximum values.

[@devendrakc8864]

Thank you for your effort sir, how can I do it in solar cell ? like input=Jsc,Voc, FF and Output =efficiency

[@DrJALAGHARI]

Thanks for the comment. Use the equation which has these values then run it for different input ranges. Then do the same procedure to train it using ANN.

[@danielkisinga6178]

kindly help me how to sort this problem/error; Output data size does not match net.outputs{3}.size. Please help with a code answer

[@DrJALAGHARI]

Which matlab version you are using?

[@muhammadashraf6645]

Sir actually what is the dataset type?is it pv voltage and pv current as an input?? Or it is temperature and irradiance??

[@DrJALAGHARI]

It depends upon your choice which parameters you are using. It makes no difference.

[@telibelakhale6667]

how make in puts and targets and what s difference between them thankyou to answer sir

[@DrJALAGHARI]

Thanks for the comment. Inputs are taken according to your system model. If you are controlling frequency then frequency will be the input and turbine valve will be the target output. Target output is the expected output that ANN has to provide.

[@telibelakhale6667]

FOR MY CASE IAM WORKING ON LANDSLIDES i have seven parametrs and every parametr have its own lass for exempl para 1 has 3 class ,param2has 7class etc how to do in puts and targets thankyou sir for answer

[@sunilbadyal4415]

nice video. but sir here i want to know how we can use "divideInd" parameter for training, validation and testing?

[@DrJALAGHARI]

Thanks for the message. You can use this testing by writing this command below the training command, and divide the data by 70,15,15% for training, testing and validation. Or you can also do testing by using nntool gui. For this purpose, you have make three variablee namely training, testing and validation data.

[@sunilbadyal4415]

Yeah sir, I have used net.divideFcn = 'divideind' And I have divide the three sets like training , validation,testing as 1:75 ,76:90 and 90:100. But I am getting error in ANN modelling as on my workspace the performance mse is different than shown on nntrain view window. Can you help me out of this error?

[@DrJALAGHARI]

@@sunilbadyal4415 I think this will be different. Because ANN gives every time different results .

[@sunilbadyal4415]

Yeah sir I know that but after we have divided the dataset using divideind it will restrict the domain of training , validation and testing data As an example we have data of 100*1 Using divideind we will fix the dataset like 1:70 for training , 70:85 for validation and 86:100 for testing . For which it will give different regression individually for training validation and testing And for output it will give total output of 100 dataset and corresponding to 1:70 we will get the our training output, similarly in case of validation and testing. Hope you will get what I m trying to say. Sir can I get your mail or other contact media so where I can easily discuss this problem??

[@DrJALAGHARI]

@@sunilbadyal4415 my email address is [e-mail]

[@DdhD-q5q]

Thank you so much

[@DrJALAGHARI]

Thank You so much

[@lampfall7915]

Good video Amy video for optimizations of pump using ANN ?

[@DrJALAGHARI]

Not yet.

[@lampfall7915]

@@DrJALAGHARI please could you help to explain about optimizations using Ann?

[@souravde6116]

Does it matter if we take I as 1000 by 3 instead of 3 by 1000. Do I require to make any changes for that in the code?

[@DrJALAGHARI]

Yes. It matters some time. Otherwise it may give error.

[@techzila5379]

Sir how we will do prediction of trigonometric functions with Ann

[@DrJALAGHARI]

Thanks for the comment. This can be done by first generating input data of the trigonometric functions by writing simple code. then use that input data to train and test the ANN output as I have explained in this video.

[@muhammadashraf6645]

what about deep neural network??it is trained in the same way?

[@DrJALAGHARI]

Thanks for the comment. I have not worked on it. But mostly procedure is same.

[@hassaniskandarani1345]

Salam doctor, I don't know a lot about neural networks, but is their a difference between the artificial NN and the adaptive artificial NN ? thank you

[@DrJALAGHARI]

Thanks for the comment. ANN basically is a system that mimics the biological neurons. Whereas adaptive ANN has the ability to change or adapt to input and output during the training stages. Basically there is no big difference.

[@hassaniskandarani1345]

@@DrJALAGHARI Okay Doctor thanks for the answer, really appreciate it

[@DrJALAGHARI]

@@hassaniskandarani1345 most welcome

[@李建霖-p2r]

very helpful !! could you make video of Three Phase Fault ,single line to ground Fault Analysis etc using MATLAB / simulink ?

[@DrJALAGHARI]

Thanks for the compliments. Ok I will try to make video on it.

[@李建霖-p2r]

@@DrJALAGHARI thanks a lot !

[@rafipermana7734]

Thanks sir! This tutorial is leading me to combinate ANN with graph, but now Im encountered an error May I send you an email for this problem? Thanks in advance

[@DrJALAGHARI]

Thanks for your message. May I know which error you are facing?

[@ArifHussain-fs5jz]

A great lecture sir.

[@DrJALAGHARI]

Thanks

[@abdulaziz7239]

Is this code is correct for the 10 random samples from range 10 to 15 : a = 5*rand(1,10)+10

[@DrJALAGHARI]

Seems correct.

[@abdulaziz7239]

@@DrJALAGHARI jazakallahu khairan

[@nurfaris1785]

sir, may i ask you? how this can be implement to fault detection in transmission line. i can understand the basic here, but is there any change for the code?

[@DrJALAGHARI]

Thanks for the comment. The ANN code remains same for all types of topics. Only input and output data will be changed.

[@snekharamesh4391]

Good morning sir,we doing one project using ANN in MATLAB for that too first part of your code is same sir

[@DrJALAGHARI]

Thanks for the compliments. Yes it will be same.

[@snekharamesh4391]

Thank you

[@2680546]

Sir! I happened to design an ANN successfully as per my requirement. Just want to know how we can determine the importance of each input variable i.e. relative contribution/effect of input variables on the value of output. Thanks in advance.

[@DrJALAGHARI]

Thanks for the comment. I think it may not be possible to know the relative importance of the input variables.

[@ashtonGM2550]

Hello Dr. Laghary, regarding the number of layers, you are saying that if we enter [3,5,1] means 1 hidden layer, but the view of the created ANN and also by comparison with other functions (e.g. net = feedforwardnet(16,'trainlm');) identifying [3,5,1] will create an ANN with 2 hidden layers with 3 and 5 nerons and 1 output and the number of inputs will be assigned by the input matrix and there is no need to enter it?

[@ashtonGM2550]

as a reference this video: kzbin.info/www/bejne/nYLJaISPr7yaoZY

[@DrJALAGHARI]

Thanks for the comment. I will check it.

[@kamranfaisal5337]

I don't get performance goal met at bottom on windows upon running this code but it shows max epoc reached , why I am unable to get result. kindly help sir regarding this. thank you

[@DrJALAGHARI]

Thanks for your message. It is ok. No issue. It is because goal set by default is 10-12 which is very large. And it is almost not possible to get this default goal. If you want to reduce the goal, then you may change it in the program. But you just open the regression curve, if R=1 or close to 1,then your ANN training is successful.

[@kamranfaisal5337]

@@DrJALAGHARI thank you, sir. how can we replace input here with values from an excel file. sir can you please help regarding that.

[@DrJALAGHARI]

@@kamranfaisal5337 if the data is small then you may copy and paste the data in the m-file. Otherwise, you may also call the excel file in the matlab. Just search the command for this. I will made another video on this topic. But it will take time

[@kamranfaisal5337]

@@DrJALAGHARI thank you, sir. as per your suggestion, I found a way to enter excel data. i have been trying to modify code in this video to accomplish a small task which is "Fault detection and classification in an electrical power transmission system using artificial neural network ". after initializing of inputs and outputs i am having error at this line : net=newff(minmax(I),[6,18,15,1],{'logsig','tansig','purelin'},'trainlm'); these are the errors i am getting : Error in newff>new_5p0 (line 285) net.layers{i}.transferFcn = tf{i}; Error in newff>create_network (line 127) net = new_5p0(varargin{:}); Error in newff (line 101) out1 = create_network(varargin{:}); Error in test (line 14) net=newff(minmax(I),[6,18,15,1],{'logsig','tansig','purelin'},'trainlm'); kindly help regarding this

[@DrJALAGHARI]

@@kamranfaisal5337 Thanks for your message. You have increased the layers by putting 6,18,15,1. Therefore, you have to assign activation function too. But you have not defined it. Just add another tansig activation function before the purelin. Then run. Hopefully, it will work

[@hattabiintisar767]

random doesn't defind in matlab 2016 ? can i find toolbooks n download it or what should i do ?

[@DrJALAGHARI]

it must be available in any matlab version. just type it correctly.

[@rameshdeepa486]

Thank you very much sir.

[@DrJALAGHARI]

My pleasure. Thanks for the compliments.

[@SikanderSingh_DR]

Sir, How will write the code for deep learning

[@DrJALAGHARI]

Thanks for the comment. You may search for it on different platforms.

[@raniaelrifai3337]

Asslam-U-Alikum Dr.J.A, Really I have question very important for me, that is about predict by Neural fuzzy, which I have followed the steps exactly and the result was perfect, but I have question that is, when I have data from a previous experiment or a study, and I applied these the steps and the results appears is well through I getting on graph ( Validation, Test, training, and All) . in this approach I have got predicted output, which I can comparative with pervious data for showing that using neural fuzzy is more accuracy than anther method, and I haven't get on optimization solve. I carried out previous paper by fuzzy logic response surface methodology by modeling and optimization corrosion rate for crude oil in pipeline. I would like to learn how after implementation by neural fuzzy by MATLAB,I can determine Error between this method and previous method, and optimization solve Thanks Rania

[@DrJALAGHARI]

Thanks for the comment. If you are comparing between ANN and other type of ANN then comparison can be made by measuring accuracy of both techniques. However, with fuzzy it may not be possible to do the comparison.

[@knowledgeclub4441]

How to extract weight from this neural ñtwork ???

[@DrJALAGHARI]

Thanks for the comment. for this purpose, kindly go through the ANN release notes from MATLAB. I dont remember exactly the code.

[@nasarbhaiahmad6176]

it is very good sir

[@DrJALAGHARI]

Thanks for the compliments.

[@aishwaryanarang2949]

Very informative :)

[@DrJALAGHARI]

Thanks for the compliments.

[@SUHAILAhmed-zr3zq]

Best way of teaching thankyou Sir❤

[@DrJALAGHARI]

Thanks for the compliments.

[@KhadizaAkter-y8w]

Sir, can u please explain ANN MPPT

[@DrJALAGHARI]

Thanks for the comment. You may go through a research paper on this topic for better understanidng.

[@abdulmoiz4450]

how to denormalize data in ANN

[@DrJALAGHARI]

Thanks for the comment. To denormalize the data, kindly use the matlab command available in ANN matlab user guide.

[@Syed-wb8lc]

Result is not shown in my MATLAB 2023a

[@DrJALAGHARI]

Thanks for the comment. What results you are getting?

[@Syed-wb8lc]

@@DrJALAGHARI Thank you for the response sir, it was a mistake which is eradicated and now its working fine. As I am using 2023a version many things are updated.

[@DrJALAGHARI]

@@Syed-wb8lc Right. Thanks

[@naveyavukkadala1811]

Can i get for simulink

[@DrJALAGHARI]

Yes. Kindly watch the fourth part of this series to learn how to use it in simulink.