Prerequisites
In order to get the most out of the post below, please check out the following blog posts before proceeding:
1. Theory: The Multi-Layer Perceptron
This is an exciting post, because in this one we get to interact with a neural network! There is a download link to an excel file below, that you can use to go over the detailed functioning of a multilayer perceptron (or backpropagation or feedforward) neural network. The video below explains the various components at a high level. If I can explain something better – please let me know using the comment section below!
If you have any suggestions on how I can improve this page, please let me know.
Clarification (2021-04-18): The neuron output function used in the excel example is a logistic function.
20 thoughts on “An MS Excel Example of a Basic MLP”
Thank you for the post. I really like it. Could you give some examples of applications of it? How one could use the obtained results?
It’s a very useful Excel Sheet, Could you help me on how to applied Sigmoid function in the sheet? Or maybe if you already done it where can I get it?
Thank you very much
Hi There
Thanks for this – very helpful for me learning neural networks and being an old hand at excel!
I think I found some tiny bugs in the formulas.. specifically, cells J5 I think should be multiplying to $F$9 (so that it reads from X2) and similar for Cell J9 this one should be multiplying to $F$4… If I’ve understood correctly in any case.
Going to try and add an extra hidden layer and/or extra nodes in the first hidden layer as bit of practice, hopefully helps me understand it more!
Cheers David
Excellent find David! I think I agree with your fix. I have edited and uploaded the updated workbook.
Great concise excel presentation! much appreciated!
Thank you very much for the video and the excel sheet.
Thank you very much for the video and the excel sheet! Also, if you were to use a sigmoid activation function, what changes would you need to make?
Great question Jordi. To use the sigmoid function the gradient calculation cells would need to change. If helpful, I can try adding a sigmoid version on a new sheet and reposting the file.
Dear Syed,
Thanks for the file and great work. On the activation function it will be great to have a comment to clarify the sigmoid from the tanh
I am new to this so your help is much appreciated.
Great video! Can you send me that excel file? thks
Thanks! You can download it by clicking the “agree to terms” and then the green button to download should activate.
Hi, I do not understand the formular for Starting row. Can you help me explain it? Thank you very much.
If you share the excel cell, I’ll try and explain.
Great example thanks. I did notice the formulas in cells F15,G15, and H15 reference column “I” and should reference column “J”. I find the MSE of the example as downloaded with 500 iterations improves from about .04 to .0003 by changing the formulas.
Great catch! Agree with the edit, file is updated. Thanks!
Dear Syed,
Thanks for the file and great work. On the activation function using Tanh is the equation =1/(1+EXP(-1.72*K4)) correct? This looks like a sigmoid function or is this a modified function.
I am new to this so your help is much appreciated.
Hi Bry – the neuron function is sigmoid, I’ll clarify and upload an edited workbook. Thanks for pointing this out!
Hello Syed,
Can you also incl. the sigmoid version into the Zip file ? then I can compare two version.
Thanks
Nice video, can you give me the excel sheet shown in the video?
Thanks
Thanks! You can download the file from the link above (under the video and right under the “captcha” link).