Data Science Stack Exchange is a question and answer site for Data science professionals, Machine Learning specialists, and those interested in learning more about the field. It only takes a minute to sign up.
Sign up to join this community
Let’s suppose I wanted to classify some input as one of three categories using a simple neural network.
The output of my network are three columns (one for each possible category I assume) with values between 0 and 1. Moreover, the single rows add up to precisely one when adding the three columns together. Is is possible to interpret the output as the probability of my input belonging to each single category?
Indeed, this is the standard interpretation of continuous classifier outputs, not only for neural networks, but for the more general case called Softmax Regression.
Thus, provided that you have used softmax activation on the final layer (in order, among other things, to ensure that your outputs indeed sum up to 1), you can interpret the continuous outputs as the respective probabilities of a particular data sample belonging to each one of your classes.
See also the discussion in this (rather unfortunately titled) discussion at SO:
How to convert the output of an artificial neural network into probabilities?
