A characterization theorem for the discrete best monotonic approximation problem

I. C. Demetriou
1990 Mathematics of Computation  
A characterization theorem is derived that motivates a procedure for generating discrete best monotonie approximations to n sequential data values, when a strictly convex objective function is used in the calculation. The procedure is highly useful in the discrete nonlinear optimization calculation that produces the best piecewise monotonie approximations to the data.
doi:10.1090/s0025-5718-1990-1023046-3 fatcat:bf24cgefingthkha4s4xevwq2u