Meta Parameters

Meta parameters are used to control the search space and the model configuration. In BSR, they are mainly defined in the theorist constructor (see regressor.py). Below is a basic overview of these parameters. Note, there are additional algorithm-irrelevant configurations that can be customized in the constructor; please refer to code documentation for their details.

• tree_num: the number of expression trees to use in the linear mixture (final prediction model); also denoted by K in BSR.
• iter_num: the number of RJ-MCMC steps to execute (note: this can also be understood as the number of K-samples to take in the fitting process).
• val: the number of validation steps to execute following each iteration.
• beta: the hyperparameter that controls growth of a new expression tree. This needs to be < 0, and in general, smaller values of beta correspond to deeper expression trees.