Skip to main content
. Author manuscript; available in PMC: 2016 Apr 16.
Published in final edited form as: J Am Stat Assoc. 2015 Apr 16;110(512):1770–1784. doi: 10.1080/01621459.2015.1036994

Table 1.

Algorithm for reinforcement learning trees

  1. Draw M bootstrap samples from D.

  2. For the m-th bootstrap sample, where m ∈ {1, …, M}, fit one RLT model m, using the following rules:

    1. At an internal node A, fit an embedded model f^A to the training data in A, restricted to the set of variables {1,2,,p}\PAd, i.e. P\PAd, where PAd is the set of muted variables at the current node A. Details are given in Section 2.4.

    2. Using f^A, calculate the variable importance measure VI^A(j) for each variable X(j), where jInline graphic. Details are given in Section 2.5.

    3. Split node A into two daughter nodes using the variable(s) with the highest variable importance measure (Section 2.7).

    4. Update the set of muted variables Inline graphic for the two daughter nodes by adding the variables with the lowest variable importance measures at the current node. Details are given in Section 2.6.

    5. Apply a)–d) on each daughter node until node sample size is smaller than a pre-specified value nmin.

  3. Average M trees to get a final model f^=M-1m=1Mf^m. For classification, f^=I(0.5<M-1m=1Mf^m).

HHS Vulnerability Disclosure