Regression
In machine learning, regression predicts a continuous numerical target from one or more features.
A simple linear predictor is
\[\hat y=\beta_0+\beta_1x_1+\cdots+\beta_px_p.\]One common training objective is mean squared error:
\[\mathrm{MSE}=\frac1n\sum_{i=1}^{n}(y_i-\hat y_i)^2.\]Nonlinear regression
Tree-based methods, kernel methods and neural networks can represent nonlinear relationships. Greater flexibility can improve prediction but also increases the need to control overfitting.
Biological use
Regression can predict quantities such as growth rate, drug response or future disease burden from measured covariates.
Key idea. ML regression focuses on accurate prediction of numerical outcomes, with performance assessed on data not used for fitting.