Novel Computational Approaches for Solving Optimization Problems

We propose to develop novel hardware-centric "initialization" technology that enables unprecedented computational power for the solution of general optimization problems. In Phase I, we will develop a hardware prototype of an "initializer" equipped with our technology, and conduct simulation studies to quantify the improvements brought by such initializer to a variety of commercial and public domain solvers so to determine the feasibility of our approach. Three mission-critical areas to the exploratory nature (called "sampling") of a general initializer are identified, namely (1) sampling speed; (2) sampling efficiency; and (3) sampling quality. Innovative algorithmic methods and architectures aimed at boosting the effectiveness of "sampling" along each one of the three areas are proposed. In particular, we propose to use hardware acceleration for speed; computational steering for efficiency; and a compensated computation of the objective function similar to that used in the sampling of analog signals, for quality. This technology has the potential to impact a wide variety of fields in which global optimization plays an important role, including engineering design, econometrics and finance, biotechnology, molecular design, scientific modeling, among others.