I know how to train a simple competitive network. Let's say I have three inputs $x_1, x_2, x_3$ and learning coefficient $\eta=0.5.$ Let's say I have two neurons $w_1, w_2$. For each input I will compute $\Vert x_i-w_j\Vert^2$ and the smallest distance will define the winner neuron. Then I will update the weight of the winner to $w_j = w_j +\eta(x-w_j)$ .

However, I am not sure how in this simple competitive network, we define the decision boundaries. For example in perceptron I know that i will draw the line $w_1+w_2-b=0$.

For example assume that I have after training:
x1=[1, 1] , class 0
x2= [-1, -1] class 1
x3 = [1, -1] class 0
w1 = [1.25, -0.25]
w2 = [-0.5, -2] .

What's the decision boundary?


1 Answer 1


Because $x$ will be assigned to the closest $w$, in the end, it'll be just like nearest neighbor boundaries, i.e. voronoi regions around each $w$.

  • $\begingroup$ is there a practical way to draw this? Because here we only have two classes, instead of n seeds (in voronoi). I cannot just take the a perpedicular to the middle of the line that connects $w_1, w_2$, right? $\endgroup$ Commented Jan 22, 2023 at 16:39
  • $\begingroup$ In case of two $w$, yes you can take the perpendicular line $\endgroup$
    – gunes
    Commented Jan 22, 2023 at 17:45

Your Answer

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

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