0
$\begingroup$

I've been reading a few Neural Networks articles for the past week and one thing that I am still trying to grasp is the dimensioning of the matrices on an ANN training. I have created a diagram (based on the example given by the same article) to try to figure it out. I have inserted the weights related to the neurons that are painted on the diagram.

I am hoping that someone can validate my reasoning, as it is one of my pain points right now.

I appreciate any help.

PS: Was this post supposed to be created on this forum or on Cross Validated? I can see that this topic appears on both.

Diagram

$\endgroup$
1
$\begingroup$

Yes. reasoning presented here is correct. Mind that $x$ input vector is a row vector (not a column one): $[x_1, x_2, x_3]$. Multiplying $x\cdot w^1$ provides you an another row vector $y_1 = [y_{11}, y_{12}, y_{13}, y_{14}]$ which represents output of neurons 1 to 4 in 1st hidden layer.
Then you multiply $y_1\cdot w^2$ and gain another row vector and so on until the outlayer. At last performing an multiplying 1x4 $\cdot$ 4x1 provides you a scalar which will be a net output after using output's neuron's transfer function on that scalar.

$\endgroup$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.