ABSTRACT
Parallel computing can provide important assistance in the design and analysis of adaptive sampling procedures, and some efficient parallel programs have been developed to allow one to analyze useful sample sizes. Response adaptive designs are an important class of learning algorithms for a stochastic environment and can be applied in numerous situations. As an illustrative example, the problem of optimally assigning patients to treatments in clinical trials is examined. Although response adaptive designs have significant ethical and cost advantages, they are rarely utilized because of the complexity of optimizing and analyzing them. Computational challenges include massive memory requirements, few calculations per memory access, and multiply-nested loops with dynamic indices. The effects of various parallelization options are analyzed, showing that, while
standard approaches do not necessarily work well, efficient, highly scalable programs can be developed. This allows one to solve problems thousands of times more complex than those solved previously, which helps to make adaptive designs practical.
