Overview#

Big Ideas#

Suupose a variable \(y\) depends on be independent variables \(x_1,\dots,x_p\). Use the notation \(\mathbf{x} = (x_1,\dots,x_p) \in \mathbb{R}^p\) to denote the vector of independent variables. The construction of a data-driven model consists of the following steps:

A data-driven model is a function \(f(x,\beta)\) and relates an independent variable \(x\) to an dependent variable $

\[ y = f(x;\beta) + \varepsilon \]

which depends on a parameter \(\beta\) where \(\varepsilon\) is the error.

When the dependent variable is continuous then we call

  • Observe samples \(y_0,\dots,y_N\) of the dependent variable and corresponding values \(\mathbf{x}_0,\dots,\mathbf{x}_N\) of the dependent variables

  • Choose a regression function \(f(\mathbf{x}; \boldsymbol{\beta})\) which depends on some vector of parameters \(\boldsymbol{\beta} = (\beta_0,\beta_1,\dots)\)

  • Choose a cost function \(C\) which measures the goodness-of-fit of the function \(f\) relative to the observed data

  • Compute the parameters \(\boldsymbol{\beta}\) which minimize the cost \(C\)

  • Analyze the regression model \(y = f(\mathbf{x} ; \boldsymbol{\beta}) + \varepsilon\) where \(\varepsilon\) is the random error representing the discrepancy in the approximation

Learning Goals#

Under construction