Right now, each segment either constrains a dimension to a particular value, or allows it to have any value. It would be good to also construct candidate segments that constrain a dimension to be within a group of values with similar properties. For example, one could run a pre-regression on just that dimension, and group the dimension values by the sign/magnitude of the coefficients.
Right now, each segment either constrains a dimension to a particular value, or allows it to have any value. It would be good to also construct candidate segments that constrain a dimension to be within a group of values with similar properties. For example, one could run a pre-regression on just that dimension, and group the dimension values by the sign/magnitude of the coefficients.