Methods for solving some optimization problems in the theory of recognition by precedents
N. N. Katerinochkina · Pattern Recognition and Image Analysis · 2014
A number of optimization problems arise when constructing various models of precedent-based recognition algorithms. One of them is the search for the maximum consistent subsystem from a given system of linear inequalities. Various approaches are proposed to solve this problem. Methods of its solution based on examining the set of nodal subsystems of a given system are developed. This may be either a complete item-by-item examination that gives an exact solution or a partial directed search that finds an approximate solution. In addition, an approximate method for solving this problem is proposed that is essentially different from the above-mentioned methods. This method is based on a number of geometric considerations.