riese@litsuns1.epfl.ch (Marc Riese) (01/20/91)
I think of `successive refinement' as follows: if a system must solve a problem in a given amount of time, or must be prepared to give its best solution at any time, then a successive refinement approach applies various problem solving techniques, that increase in terms of time requirements and quality of solution, until the solution is requested, or the allotted calculation time expires, or all the applicable techniques have been tried. I have recently seen and heard references to the same idea, but using a different name ("progressive reasoning", "anytime algorithms",...). Can anyone tell me: 1. Who came up with the idea? Maybe it's so obvious that it shouldn't be attributed to anyone. 2. Are there real differences corresponding to the different names given? No hair-splitting please. Offhand, I would consider any discussion about the solution algorithms used as hair-splitting. Marc Riese Swiss Federal Institute of Technology