How to use an already trained perceptron?

Question

Python

I already picked up the weights, but I don't know how to use them for the new case. I thought about calculating the dot product of new cases with weights (forward propagation), but the results were wrong.

Answer 1

Already understood. You can calculate the result of a new case like this: sigmoid(dot(new_case, weights)), where dot() is the dot product of the input data with weights.

Answer 2

If your network predicts poorly on new data, then you need to train further.