Specification of a Contraction Algorithm
Neil W. Tennant · Oxford University Press eBooks · 2012
The chapter explain the concept of a greedy algorithm, and provide four successively less-greedy versions of a contraction algorithm. Here, the earlier work familiarizing the reader with dependency networks and their coloration conventions is key. The chapter discusses further possible sophistications of our approach, which would involve relaxing some of our simplifying assumptions: (i) making use of information about relative entrenchment of beliefs; (ii) being able to give up transitional steps among beliefs, in addition to beliefs themselves; and (iii) weakening the degree of support that the premises of a step afford its conclusion, so that it is probabilistic rather than apodeictic. The fourth version of our algorithm addresses (i).