The theory of optimization by accuracy of finite-dimensional methods for solving various classes of linear problems is presented. Considerable attention is paid to optimal approximations of solutions of integral and integro-differential equations.