Doubt related to a question from "Quiz: Sigmoid - Data and Tasks"

Here, the correct answer given is that, we can’t predict the amount of oil that can be mined from a particular point in an ocean using the sigmoid neuron, we can only predict the probability of finding oil at that point.

My doubt is that, if we normalise the value of amount of oil to a value between 0 and 1, then why won’t sigmoid neuron help in predicting that value??

Hi, It’s because of the following reasons:

  • Normalising will give us an output between 0 and 1, which is not the desired range for amount of oil.
  • Sigmoid gives us probabilities as output.
  • Sigmoid is one of the most basic models, and the data would be underfit in most cases.