A Brief Survey of Knowledge Aggregation Methods.
Abstract
This paper discusses several iterative knowledge aggregation methods. Such methods are used to choose one of a finite set of labels about each of a set of objects. First, a stimulus is analyzed locally at each object, yielding an initial state which assigns a weight of the evidence from that analysis to each of the labels. The methods continue as a sequence of trials. On each trial two events occur. First, new evidence is gathered either externally or using internal compatibility constraints on the current state. Second, the current state and new evidence are combined, resulting in a new state, which becomes the current state for the next trial. This method iterates until sufficient confidence in a single label at each object is achieved. In this paper, several such methods are reviewed and compared in terms of the form of the state space, the type of evidence which can be represented, and the efficiency and convergence properties of the methods as a whole. Keywords: stochastic method; relaxation labeling; stochastic relaxation.
Document Details
- Document Type
- Technical Report
- Publication Date
- Sep 01, 1985
- Accession Number
- ADA178957
Entities
People
- Michael S. Landy
- Robert A. Hummel
Organizations
- New York University