37. Overview of scientific modeling material

The material here supplements the Bayesian methods for scientific modeling introduced in Bayesian methods for scientific modeling with a more traditional viewpoint.

• Overview of modeling address scientific models in general and regression analysis in particular. Comparisons are made between parametric and non-parametric models and between linear and non-linear models.

• Linear models gives a complete treatment of conventional (as opposed to Bayesian) linear regression.

• Mathematical optimization describes optimization schemes, which are of particular interest for minimizing loss functions of neural networks. In particular, gradient descent is covered in some detail, including stochastic and adaptive algorithms (see also Gradient-descent optimization).