# Need some Guidance with Excel ANN

#1

Hey Everyone, glad to be a part of the community.

The content on this website is fantastic. I’ve tried learning more about machine learning and AI on other platforms/resources and none of them have explained it quite as well. Thanks!

Anyway, I did have a question about one of the articles I read, this one in particular:

A great breakdown of how ANNs work. However I am lost on one step. In the visual section on step 10, I can’t quite figure out how the formulas should be. Can anybody post an image of exactly how the formulas should be broken down, or perhaps break this step into more steps so it is more clear?

This is the formula I cannot quite figure out :

wh = wh + matrix_dot_product(X.Transpose,d_hiddenlayer)*learning_rate

What it’s telling me is that X should be transposed in the Matrix Multiplication. Which is necessary because you only multiply rows by columns, and vice versa. However, which part of X and which part of d_hiddenlayer gets multiplied by each other?

If someone could break this step into more steps that would be helpful.

Thank you.

#2

Hi @rich90,

Suppose your X matrix is given by:

And the delta at the hidden layer (d_hiddenlayer) as calculated in step 9 is:

Now first you take the transpose of X matrix, which will be:

Finally, you perform the matrix multiplication and calculate matrix_dot_product(X.Transpose,d_hiddenlayer), i.e. each row of X.transpose will be multiplied with each column of d_hiddenlayer:

After calculating these values, you will get a matrix like:

Now you multiply this with learning rate and add the previous wh value to get the final wh value which will be:

Hope this helps!!

#3

This was in fact very helpful. Thank you for breaking it down in a visual way. Very much appreciated.