# Target analysis

This is a metaheuristic to solve global optimization problems, notably combinatorial programs, using a learning mechanism. In particular, consider a branch and bound strategy with multiple criteria for branch selection. After solving training problems, hindsight is used to eliminate dead paths on the search tree by changing the weights on the criteria: set $LaTeX: w > 0$ such that $LaTeX: \textstyle w V_i \le 0$ at node $LaTeX: i$ with value, $LaTeX: \textstyle V_i,$ that begins a dead path, and $LaTeX: \textstyle w V_i > 0$ at each node, i, on the path to the solution. If such weights exist, they define a separating hyperplane for the test problems. If such weights do not exist, problems are partitioned into classes, using a form of feature analysis, such that each class has such weights for those test problems in the class. After training is complete, and a new problem arrives, it is first classified, then those weights are used in the branch selection.