Hi,
When we normally implement the Perceptron model in scikit learn it doesn’t perform any probability estimation. In the programming exercise in week 2, do we actually calculate a probability or we classify if the linear combination exceeds a threshold ?
The output of sigmoid is not technically a probability distribution, but the values behave like probabilities. They are between 0 and 1 and if we interpret a value > 0.5 as predicting “yes” for a given input, then we can use that in combination with the “log loss” or “cross entropy” loss function and it evaluates the result in a way that can be used to drive the training of the coefficients to give a better answer by using back propagation.
Thank you for your response 