Finding the function mapping in supervised learning is essentially an optimization problem that adjusts model parameters by minimizing the loss function (difference between predicted and true values).