Merge pull request christophM#348 from shokru/patch-1

Update 04.7-interpretable-rulefit.Rmd

manuscript/04.7-interpretable-rulefit.Rmd

@@ -225,7 +225,7 @@ Since RuleFit uses Lasso, the loss function gets the additional constraint that
 
 \begin{align*}
 (\{\hat{\alpha}\}_1^K,\{\hat{\beta}\}_0^p) & = & argmin_{\{\hat{\alpha}\}_1^K,\{\hat{\beta}\}_0^p}\sum_{i=1}^n L(y^{(i)},f(x^{(i)}))  \\
- & + & \lambda\cdot\left(\sum_{k=1}^K|\alpha_k| + \sum_{j=1}^p|b_j|\right)
+ & + & \lambda\cdot\left(\sum_{k=1}^K|\alpha_k| + \sum_{j=1}^p|\beta_j|\right)
 \end{align*}
 
 The result is a linear model that has linear effects for all of the original features and for the rules.