-1

I'm programming a simple one layered perceptron. For example, I have 3 neurons at the first layer, 2 at the second, and 2 at the output layer. I have to solve a binary classification problem. This way I have 10 weights. But I want to visualize the function that I get from this weights. E.g I want to plot function y = w0 + w1*x So, the question is, which w0 and w1 are proper for this purpose? enter image description here

Rahul
  • 1,330
  • 2
  • 13
  • 25
  • 1
    I'm not sure if it's relevant, but this JS NN library has an option to convert your perceptron to a standalone function. https://github.com/wagenaartje/gynaptic – Thomas Wagenaar Mar 19 '17 at 11:15

1 Answers1

0

You do not get a single function. You can derive the decision function from the weights, but it's not a trivial combination of any pair of inputs; rather, it's that composition of the two layers of functions -- and then you subtract the values computed for class1 and class2; positive is one class, negative is the other.

If you want to plot the input weighting function for a particular perceptron, then you need to use the obvious w values, but I don't think this is what you're asking.

Prune
  • 72,213
  • 14
  • 48
  • 72