Dimensionally Independent Response Surface Method

Dimensionally Independent Response Surface Method (DIRSM) extends multi-objective optimization to large scale problems. DIRSM works with dozens, hundreds, or even thousands of design variables, with no increase in the number of evaluations necessary to find Pareto optimal points. DIRSM dynamically creates an approximation formula based on a small initial number of evaluations, and automatically selects the most significant design variables. Unlike other Response Surface Methods, DIRSM automatically recognizes and utilizes only the most significant design variables. DIRSM uses a constant number of evaluations (typically 4-6) to build its Response Surface, rather than other methods where the number of evaluations grows exponentially with the number of design variables.

DIRSM not only focuses on the most significant variables, but dynamically adjusts the formula as the optimization process runs. So rather than a static formula that must calculate a coefficient for each input parameter, DIRSM utilizes a dynamic formula that focuses on the most significant input parameters that are found as evaluations are done. The result is that as the number of input parameters grows, the number of evaluations required by DIRSM is constant. For problems with large number of input parameters, DIRSM provides both CGM and PNM algorithms with the capability to generate Pareto Points with an exceptionally small number of evaluations.

See A Demo White Paper

Dimensionally Independent Response Surface Method (DIRSM) maintains a constant response time independent of the number of variables. DIRSM (in blue and green above) other algorithms require orders of magnitude more evaluations as the number of variables increases.