ABSTRACT
Recall that given an optimization model we are also presented with, typically, a large space of decisions for the model. There is by definition one decision for each possible setting of the decision variables. Our Simple Knapsack 100 model has 100 binary decision variables, so there are 2 = 1.2677 × 10 decisions associated with the model. Our GAP 1-c5-15-1 model has 15 decision variables, each of which can take on 5 values, yielding 5 = 3.0518 × 10 decisions in all. Finally, in our Philadelphia 2010 Census Tracts model there are 1.6625 × 10 decisions possible when choosing 10 hubs out of 383 tracts. Our examples have millions and millions of decisions. They are large enough to be good for illustrating the points we wish to make, small enough to be effectively tractable, and towards the low end of complexity for industrial problems.
