0
$\begingroup$

enter image description here

Starting with a).

For the first unit: 0 * x1 + 1 * x2 + 1 > 0 (0, because the threshold is 0) which is the same as x2+1 > 0.

For the second unit: x1 * 1 + x2 * 0 + 1 > 0 (0, because the threshold is 0) which is the same as x1 + 1 > 0.

For the third unit: x1 * 1 + x2 * 1 + 1 > -1 (-1, because the threshold is -1. I might have misunderstood this since the assignment states "Circles indicate threshold units with threshold at zero...") which is the same as x1 + x2 > -2.

I then draw those lines. Is it correct so far? I can't however figure out how to solve b) How do I take the bias on the output node into account?

This is the given solution (but I don't understand how they got it):

enter image description here

Thanks!!

$\endgroup$
8
  • 1
    $\begingroup$ I think you mixed intercept and threshold. The threshold is set to zero (read the text). The intercept is 1 * 0.0 for the first two nodes and 1 * (-1.0) for third one. This leads to different inequalities. $\endgroup$
    – Broele
    Mar 10 at 13:05
  • $\begingroup$ By the way: this looks like you are asking to do your homework. I know such kind of networks mainly from text books and for educational purposes. $\endgroup$
    – Broele
    Mar 10 at 13:07
  • $\begingroup$ Thanks for your reply! I thought I was supposed to use this equation Σwjxj+bias=threshold for the nodes. But I guess that isn't the case since I get the wrong answer if I just plugg in bias = 1 and threshold = 0, and I don't see anything about any 'intercept'. What equation should I use? I feel a bit lost and I'm having a hard time figuring out how it all fits together. @Broele $\endgroup$
    – kim120
    Mar 10 at 13:27
  • $\begingroup$ Bias and intercept mean the same, here $\endgroup$
    – Broele
    Mar 10 at 13:50
  • $\begingroup$ Ok thanks! But what do the numbers next to the filled in circles stand for? I thought those were the bias. @Broele $\endgroup$
    – kim120
    Mar 10 at 14:01

1 Answer 1

0
$\begingroup$

What you assumed were thresholds appear to instead be weights on the edges from the passthrough nodes. At least, that makes the lines consistent with those in the plot.

I gave some hints to the same problem at Draw(by hand) the decision boundary of an neural network ; for the shading, note the output of each hidden neuron in each region and plug those into the output neuron.

$\endgroup$
3
  • $\begingroup$ Thanks for your answer! But if those numbers I assumed were thresholds actually are weights on the edges from the passthrough nodes, then what are the '1.0's closer to the output node? I thought the '1.0's were the weights... @Ben $\endgroup$
    – kim120
    Mar 10 at 13:47
  • $\begingroup$ These are the weights for the output of the hidden layer when uses as input for the output layer. $\endgroup$
    – Broele
    Mar 10 at 13:54
  • $\begingroup$ @kim120 those are the weights for those edges, giving the output as a linear combination of the (post-activation) output from the hidden neurons. $\endgroup$
    – Ben Reiniger
    Mar 10 at 14:17

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge that you have read and understand our privacy policy and code of conduct.

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