A note on the block conjugate gradient method of O'leary

In this paper we show that many important algorithms of conjugate gradient type may be regarded merely as particular versions of the block method proposed by O'Leary, and argue that the use of this method as a theoretical tool will result in simplifications of great value to the theory. This argument is supported by the inclusion here of a new method whose existence the author had long suspected but had been unable to find prior to exploiting this simplification