How does the decision boundary affects the selection of parameters?
I mean, for different decision boundaries set by us (for e.g. 0.5 or 0.7 or 0.8 etc.) will we get different set of parameters?
If yes, then how does this reflect in the cost function equation?
Hi @Thala, if you are asking about the logistic regression, the decision boundary won’t affect the optimized weights because we don’t consider the decision boundary in the optimization process. Remember we use the log loss -y\log{(f(x))}-(1-y)\log{(1- f(x))} and the log loss considers the model prediction as is - a probability value ranged between 0 and 1, instead of a binary prediction.
Raymond
So the logistic regression will return a value between 0 and 1 lets say 0.7. If i set my decision boundary 0.8 then i will consider the returned value as 0(since 0.7<0.8).
Am I interpreting correctly?
Hello Thala, yes, your interpretation for how to make predictions using your trained logistic regression model is correct!
Raymond
1 Like