When generating multiple numerical columns each of them will fill its one dimensional space. However if you want to cover a high dimensional space it is important that there are no uncovered areas. This can be achieved by quasirandom sequences. All dimension will be in the interval from 0 to 1. By increasing the number of data points (with a bigger input table) the "grid" of points in space gets finer. There are currently two methods provided: the Halton sequence generator and the Sobol sequence generator. Please note that the Halton generator only support up to 40 dimensions and the Sobol up to 1000.
The generation is performed with the Data Generators as provided by Apache Commons Math.