Asymptotic theory of selection by relative rank with high cost

Asymptotic theory of selection by relative rank with high cost

0.00 Avg rating0 Votes
Article ID: iaor19951492
Country: Japan
Volume: 36
Issue: 4
Start Page Number: 243
End Page Number: 253
Publication Date: Dec 1993
Journal: Journal of the Operations Research Society of Japan
Authors:
Keywords: decision theory, combinatorial analysis, simulation
Abstract:

Selection from among n objects by relative rank with no recall - the ‘secretary problem’ - in the asymptotic case when equ1 is considered, assuming that k, the cost ratio, becomes very large and equ2 is kept to be a finite constant. It is known that in the ‘low cost’ case where K=kn is kept to be a non-negative finite constant, the expected number of observations comes out to be O(n) and the expected absolute rank of the selected object comes out to be O(1), and that in the ‘medium cost’ case where kis kept to be a positive finite constant, both expected values come out to be equ3. Here the former comes out to be O(1) and the latter O(n). teh graph of total ‘loss’ vs. K looks like a continuous broken line, while those of ‘expected cost of observations’ and ‘expected absolute rank’ have jumps at many points. As k roaches 0, the curves approach smooth ones corresponding to the ‘medium cost’ case.

Reviews

Required fields are marked *. Your email address will not be published.