Machine Learning Idea – Non Linear Probability Curve? With Respect To The Number Of Training Samples

Since machine learning models can be non linear. Would not the predicted probability of label output of the model change in a non linear fashion. With respect to the number of training samples. I will try to plot such a curve.