f defined on the set T of matrices with values in the set of matrices
is said to be semi-convex if X, Y is-an-element-of T, X less-than-or-e
qual-to Y implies that lambdaX + (1 - lambda) Y is-an-element-of T, 0
< lambda < 1, and f(lambdaX + (1 - lambda) Y) less-than-or-equal-to la
mbdaf(X) + (1 - lambda)f(Y). Here X less-than-or-equal-to Y means that
every element of Y - X is nonnegative. In this paper, a number of dif
ferent inequalities are given for semi-convex functions of matrices. (
C) 1994 Academic Press, Inc.