Create a Breast Cancer Detector With Only 2 Data Points

One of the most common dramas experienced by data scientists is the uncertainty that the available volume of data will suffice for model building. Challenging such insecurity, I've decided to make a Machine Learning algorithm fitted with the lowest possible amount of instances. After many reflections, I've found that working with a binary classification task would be an interesting idea. Since I wished to use the minimum amount of data, the final model needed to be fed with only two instances, each one belonging to a specific category. The data points selection would be performed with K-Means.