Machine learning builds mathematical models from data containing multiple attributes (i.e., variables) in order to predict some variables from others. For example, in a cancer prediction application, each data point might contain the variables obtained from running clinical tests, whereas the predicted variable might be a binary diagnosis of cancer. Such models are sometimes expressed as linear and nonlinear relationships between variables. These relationships are discovered in a data-driven manner by optimizing (maximizing) the “agreement” between the models and the observed data. This is an optimization problem.