To the theory of the alternating triangle iteration method

We construct a new class of versions of the adaptive alternating triangle method whose optimization requires no a priori spectral information in contrast with the approaches known so far. The same estimate for the convergence rate is preserved as in the presence of a priori information.