Would you please solve and describe the proper answer for 1 and 2? This is for Machine Learning concept

Where we are doing supervised learning, we have mostly assumed a deterministic function. Imagine instead a world where we are trying to capture a non-deterministic function. In this case, we might see training pairs where the x value appears several times, but with different y values. For example, we might use attributes of humans to the probability that they have had chicken pox. In that case, we might see the same kind of person many times but only sometimes they may have had chicken pox. We would like to build a learning algorithm that will compute the probability that a person has chicken pox. So, given a set of training data where each instance is mapped to 1 for true or 0 for false: 1. Derive the proper error function to use for finding the ML hypothesis using Bayes' Rule. You should go through a similar process as the one used to derive least squared error in the lessons. 2. Compare and contrast your result to the rule we derived for a deterministic function perturbed by zero-mean gaussian noise. What would a normal neural network using sum of squared errors do with these data? What if the data consisted of x,y pairs where y was an estimate of the probability instead of 0s and 1s