Step-by-Step Beginners Tutorial: How to Train an Artificial Neural Network with Matlab

The step-by-step detailed tutorial walks you through the process of building, training, and using an artificial neural network (ANN) from scratch using Matlab. As a demo application, we use this model to predict the risk of fire in a forest as a function of temperature, relative humidity, and wind speed.

[@Luis-ty9dq]

This is just pure gold. Way better than the professors that I pay thousands of dollars for in college. Thank you so much for sharing your knowledge in such an understandable way.

[@mansourehrezapourian1404]

Many many thanks. I really understood NN from the scratch, thanks for spending time. Your explanation was so straightforward and so understandable. I can say that was one of the best training videos. Thanks again Be safe and healthy

[@AbidAli-wp2fs]

I can’t thank you enough for this wonderful masterpiece. I was developing an ANN for engine net brake torque estimation ( this info is required for Transmission Control unit) for gear shift quality. I learned from the best 🙇‍♂️ ❤

[@LuisAcosta-cf8kn]

Words cannot describe how relieved I am; your video helped me put together and understand 4 weeks worth of looking into NN for multiple numerical inputs, thank you so very much!!!

[@nazirahmohdrazali1179]

Love the way you explain for a newbie like me..

[@samuelgbashi8969]

Thank you so much I enjoyed the video very well. It is one in a million You are a rare teacher.

[@user-zq4cw6wn1o]

This video is a gold mine

[@SalahEL33]

Merci beaucoup cher Mathieu! You're the best keep it up bro

[@lorenzomason3345]

Man i really want to thank you for your precious job. No video out there about feedforward regression problem with fitnet, good job

[@aowalsalis6506]

Thank you so much for the wonderful presentation.

[@yonasabebe]

Very nice piece to undertsand how to build ANN. Thank you!

[@rancreation2012]

Tahnk you for amazing video...

[@chikopheidris394]

Quite impressive. I have met the right teacher!!

[@ManMo353]

Thank you for your wonderful presentation It was really helpful

[@stanage_]

Loads of thanks!! This was very descriptive and informative. Please can you do a video on solving a similar problem without using an inbuilt a NN function, but instead using back propagation. This will be really helpful, thanks.

[@elprofessor8922]

Thanks for this video. It was really helpful for me. Nice and pleasure English language with French accent. Greetings from Poland 🇵🇱

[@kjyu]

Superb explanation, very informative!

[@ashrafm.s2876]

Wow, It's a very wow helpful tutorial.... Thanks.

[@shashankgupta2749]

Thanks a lot. This video is quite helpful!!

[@ashenafibelihu1123]

thanks a lot! really worthy & helpful vedio to understand ANN from scratch!!

[@hichchad7309]

Thank you sir 🙏🙏 .. you're my saviour ✌

[@user-pk4ji8wm8s]

great video, looking forward one for custom NN

[@mgh5913]

That was great. Thank you a lot. 🙏

[@muhammadneamatullahkhan6210]

Thanks a lot for such excellent explanations

[@xiaowen19920501]

Thanks a lot, it do help me to understand how to make it work by using Matlab

[@fkratlatif495]

Many thanks, for Sharing this

[@phunguyentrung5796]

Thanks for the video, it helped me alot

[@riokleva3497]

Very helpful video, thank you

[@DrMuhammedAliErbir]

You Are EXCELLENT, Thank You

[@minaishac1390]

Many Thanks for this video. Can you please share with us the CSV file of the data set ?

[@salehabuzir5642]

Thanks a lot! This is a masterpiece! It helped me a lot

[@ah89971]

You are hero. Thank you so much

[@javadrahmannezhad9908]

Though without the data file it is hard to follow, the explanation was great.

[@tonmoysikder9266]

wow, it's very helpful tutorial

[@nagannasollapure4771]

Sir, it is really worth to watch this video, specially for ANN how it works in different scenario with various hidden layers, thank you for sharing your knowledge. where I will get data set for fire to check with different scenario

[@yl96]

I fell asleep a few times during your class, but the ANN is clearly introduced and performed anyway.

[@xylophoneize]

brooo 😀😀

[@nafaszareee6871]

Very good. thank you for vodeo training. i neaded this is eduction. thankes thanks:)

[@anupamabose4784]

Excellent.

[@whteo2443]

Hi sir, thank you so much for the informative video! However, I would like to ask how to fix the weights and biases once we determined the best representable model for the dataset? And is there a way to use this ANN model as fitness function of Genetic Algorithm? Hope for sir's reply. Really appreciate it, thank you.

[@mehbubulhaquebhuiyan4746]

Hi, this is a great video. May I know how we can extract the predicted data points in the form of a table?

[@salsal7309]

Thank you so much, I have a problem in validation, Matlab says (Reference to non-existent field 'ValInd') with all validation formulate

[@aizazahmed6406]

Thank you so much for detailed video. It was really helpful in understanding whole code. As a beginner in MATLAB NN it really helped me.

[@dr.mahanteshm.nadakatti3472]

Excellent explanation indeed!. Each step clearly explained. I'm a novice in the field of AI & ML. I could develop similar model using this explanation, without tampering much with either the name of the column headers or with the number of columns. However, after successfully developing the model, I got this question as to why the readily available ANN app in MATLAB was not used for developing the model or solve it. Can anyone clear my doubt please?

[@ilariavestale9842]

Hi, thanks for the explanation but there are ways to understand: - how many hidden layers and how many hidden neurons to use? - network type, traning fuction, adption learning fuction, performance fuction and transfer fuction? thank u so much

[@NissabouriSalah]

Thank you very much 👍👍👍

[@ermiaswubet5339]

I am doing research work on optimization using ANN using three parameters with three levels(L9 orthogonal array) with a single response/output, please help me how to get the optimum parameters

[@karimfarag5321]

good job

[@stephentete1211]

hello, nice and great presentation My Question is that: Why wasn't the output (y) initially normalized?

[@realarfin.raihan134]

It's awesome... Thank you so very much for this amazing video... I was searching for such a content for a long time... Thank you for making it simple and easy...

[@mustapham6890]

Thank you , thank you from the bottom of my heart. That really helped me. You are genius. Your detailed explanation for every step waaw. May I ask you that you do an example using " RNN " please :) Again my sincere appreciation for the effort placed in that video. God bless you

[@MT-fp1bj]

Can MAPE (Mean Absolute Percentage Error) use instead of RMSE?

[@realengineering2178]

Thank you very much for doing this. However, I have one question for you. In MATLAB, if you use fitnet it automatically normalizes the data using mapminmax, basically both input and target. In that case, i feel, you normalize it twice. Can you please respond? Thanks again

[@fkratlatif495]

However, I have no negative value in my output, after training the NN I get a negative value, can you please help, how can I avoid getting negative values? I have tried using the sigmoid function, it did not work.

[@maliks.a.9044]

Thank You for the simple explanation. Can You share the csv file please

[@kukh1370]

thank you for informative video, it is very helpful. but i have question, after modeling how can we use the model? is it possible to export the model we made? . what fuction should i use to use the model we made?

[@nagannasollapure4771]

Sir, based on which method/algorithm dataset output calculated...by fusing the data like temp/wind and humidity

[@mohamedelibrahimi8161]

i have a problem in understanding the reversing of the normalisation (of input data) automatically done by Matlab. I have followed the steps discribed in matworks.com website but it does not works (the analytial function is to far than the value evaluated by net(..)). if you can help me, this is my code: entree=xlsread('O:\stage\database1_sans_considere.xlsx',4); % size=[24960 8] sortie=xlsread('O:\stage\database1_sans_considere.xlsx',5); % size=[24960 1] n=6 ; net=fitnet(n); [net,TR]=train(net,entree',sortie'); %evaluation of the analytical function x=[ 1100 , 1155 , 10 , 1 , 0 , 0.7 , 343.7508 , 1.00017 ]; b1 = net.b {1}; b2 = net.b {2}; IW = net.IW {1,1}; LW = net.LW {2,1}; a1=IW*x'+b1; y=purelin(LW*(tansig(a1))+b2); % comparison with net(x')~ 10.37 error=abs(net(x')-y);

[@AbidAli-wp2fs]

I will appreciate if you can perform optimization to get the optimal number of hidden layers as you did for neurons.

[@mdsazzadurrahman4811]

I hope I am not asking too much, but would it be possible to share the dataset? Then I would have performed the tasks as shown in the video. For a beginner, it would be of great help.

[@kevinkabila5996]

How to optimize an Ann with genetics algorithms in matlab

[@mahvishjan3706]

Sir what if output is a string how can i pass that as an argument

[@add-wisehaalathakikatkeanu5342]

How to save the NN after its creation; hw to use it later on for ex for cross-validation

[@user-ph2gx8gx7v]

Sir, I can use the same code for bearing fault classification.

[@user-ms9xt7dy5l]

Thank you, Prof. Is it convenient for you to share the training and testing data used in this example?

[@saijagannadhnadimpalli4162]

can you please show how to predict using this ANN model

[@larryfocus]

Great tutorial! The demonstration is super clear and well organized. One question though - at 1:02:40, you mentioned ideally the same model need to be trained several times to get more accurate results; but I thought each run is independent of one another, so why should we expect better results (or smaller RMSEs) with more runs? And how should the results be reported if each run gives a different set of RMSEs? Thank you.

[@mahmoudzemzami3570]

I had the same question.

[@preetamchatterjee1723]

I might be wrong but my guess is that since each run took 70% of the training data randomly out of the the 500 datapoints, the input data is hence different in each case and so every iteration will give a different RSME. Same logic applies for the validation dataset. The RSME values can increase or decrease for the subsequent runs. I think he told us to perform multiple iterations so that we have an idea of what the mean error is turning out to be.

[@leewenzhi4119]

Very thank you. Can you please share the CSV file and the code.

[@asifhasan359]

First of all thank you very much for this video. It really helps me alot. Could you please clear my one doubt? like how should i change the activation function to ReLu in MATLAB.

[@0713059059]

Thank you. Can you please share the CSV file. It will be easy for us to follow the tutorial.

[@samyakbahuguna6014]

where you able to find the csv file?

[@devanshverma6189]

do you have csv file ?

[@vikasmishra4871]

how to use "fitnet" in octave...?

[@shabinjoby5501]

Hi I am getting an error in the main code “Target T is complex”

[@wgpark9189]

thanks a lot PARISlab@UCLA recently i just started to learn about ANN. it was really cool video. Could you give me the CSV file? i'd like to do this tutorial

[@pierre5821]

Hi, thanks, great video, where could we find the initial data ?

[@devanshverma6189]

Can someone please share this code file in csv,txt,or matlab format anytype of file format would be great help for understanding the concept .

[@treynou1384]

Can i still follow this if i have 2 output?

[@lukasevsek6546]

How do you set the other Hyperparameters of ANN? For example: Learning rate type, Initial learning rate or Solver type?

[@mohamedelibrahimi8161]

Thank you very much for this amazing video. Please, how we can find the final optimized values of biais bi and weights wi and the final analytical model (prediction function) using Matlab?

[@whteo2443]

Hi. May I know if u have found the way of obtaining the simulation function of ANN model in the end?

[@taracat862]

says "Unable to perform assignment because the indices on the left side are not compatible with the size of the right side." when ı try to find rmse in the loop

[@jonathankabasele985]

merci matthieu, tu merite une biere

[@ajax6262]

I need immediate help, i got stucked in the part of optimizing the neural network 56:00. It told me that the size in the left and right have different number of elements. To give a headstart, I have 3 outputs. What should i edit in the upper part of codes for it to work? The video only explained a one output model but mine's 3

[@realarfin.raihan134]

Thank you so much for such an informative video.... Can someone kindly help me to know why there is no data set for test set...?? Does this effect the overall optimization

[@GundoganFatih]

Thank you for sharing the tutorial, it would be nice for us to have CVS file to use the same codes using our computers. Can you please share it also?

[@harshitham765]

Excellent tutorial sir, thank you so much. what are the other ways of normalizing the data ? without using log function .

[@zhuokim1802]

guys,how can you get the csv. ?

[@tesfalemabraham6151]

This is nice and detailed tutorial, thanks. But how can we apply the trained Ann model for the same input data but different time periods (it could be future period)?

[@techzila5379]

sir can you make vblog on fault tolerant control system?

[@qqn4531]

La France représente !

[@solajegs]

Does anyone have the m-file of this lesson?

[@dimasrizky1413]

can you share the CSV file for us?

[@aimanfahmy3734]

32:34 neural network explanation

[@meryemzarboubi5682]

can you please give us the link to download data !!! sooos

[@Vishal360]

Can u provide data set?

[@zuhalhamza3235]

How to get the matrix if I put my matrix don't come to the Matlab

[@ahmadkelixo7243]

permissio to learn sir

[@nagannasollapure4771]

please share the CSV file and relative code

[@prakashv104]

[fs,history]=sequentialfs(classf,X_train,y_train,'cv',c,'options',opts,'nfeatures',2); What is the error in this line ? Can anyone help to figure it out?

[@eshetieberhan7782]

can you share us the data

[@medetzhangaliyev2151]

how to add 2 or more inputs to the ANN model?

[@polatburakcan]

If you ask how to add another features of ann model such as weights, bias etc., it is not explained in this video. You can check MATLAB’s guide on ANN toolbox. If you ask inputs for the output, he uses already 3 inputs data.

[@vasachisenjubean5944]

May be i am blind coz i dont see the dataset

[@edbanbs]

Its the same plot, you used de same plot to 7 neurons to 1 neuron

[@edbanbs]

sorry Im wrong, later you corrected it