Decision-making with incomplete information
We must often make decisions in situations such as routing, scheduling, compiling, and resource allocation, without crucial information about (respectively) the terrain topography, the execution times, the run-time environment, and the future requests. Information may be unavailable because of its temporal or distributed nature. A reasonable way to assess the effectiveness of a decision rule in such situations is to compare its outcome to the ideal optimum solution, attainable only if we had complete information. An algorithm is considered effective (or “competitive”) if its performance is a multiplicative constant away from the ideal solution. We review some recent and on-going work on this active area.