Showing papers on "Decision tree model published in 1988"

Decision tree design from a communication theory standpoint

Abstract: A communication theory approach to decision tree design based on a top-town mutual information algorithm is presented. It is shown that this algorithm is equivalent to a form of Shannon-Fano prefix coding, and several fundamental bounds relating decision-tree parameters are derived. The bounds are used in conjunction with a rate-distortion interpretation of tree design to explain several phenomena previously observed in practical decision-tree design. A termination rule for the algorithm called the delta-entropy rule is proposed that improves its robustness in the presence of noise. Simulation results are presented, showing that the tree classifiers derived by the algorithm compare favourably to the single nearest neighbour classifier. >

Decision trees and downward closures

Abstract: The separation of small complexity classes is considered. Some downward closure results are derived which show that some intuitively arrive at results that were published previously are misleading. This is done by giving uniform versions of simulations in the decision-tree model of concrete complexity. The results also show that sublinear-time computation has enough power to code interesting questions in polynomial-time complexity. >

On the communication complexity of graph properties

Abstract: We prove t(n log n) bounds for the deterministic 2-way communication complexity of the graph properties CONNECTIVITY, s-t-CONNECTIVITY and BIPARTITENESS (for arbitrary partitions of the variables into two sets of equal size). The proofs are based on combinatorial results of Dowling-Wilson and Lovasz-Saks about partition matrices using the Mobius function, and the Regularity Lemma of Szemeredi. The bounds imply improved lower bounds for the VLSI complexity of these decision problems and sharp bounds for a generalized decision tree model which is related to the notion of evasiveness.

Information tree: a model of information flow in complex organizations

Abstract: A model of the information flow in information processing organizations is presented. The purpose of the model, called an information tree, is to capture the essential information activities that occur as an organization performs its tasks. An information processing organization transforms requests (inputs) into responses (output) through a set of procedures or processing rules. Information trees model such transformations as changes in the information state vector of various entities in the organization. The information state vector is a dynamic quantity similar in concept to the state of a system. The information-tree model has several analytic uses. One of its most important uses is an an aid to reorganizing offices as new technology is introduced. In particular, there is a strong structural linkage between the information-tree model and the problem of optimal control of dynamic systems. The linkage is demonstrated by use of dynamic programming to solve the reorganization problem using the information tree model. >

Measures of complexity : proceedings of the conference held in Rome, September 30-October 2, 1987

Abstract: Complexity and forecasting in dynamical systems.- On Complexity.- Boolean networks which learn to compute.- A dynamical learning process for the recognition of correlated patterns in symmetric spin glass models.- Neural networks that learn temporal sequences.- Hierarchical diffusion.- Random walks, random surfaces, and complexity.- Complexity in large technological systems.- An introduction to the theory of computational complexity.- Measures of biologically meaningful complexity.- Complex systems, organization and networks of automata.- Complexity in ecological systems.

A statistical model for the search problem

Abstract: Searching a file of records or list of values for a particular member of the file or list is a fundamental problem in computing. Unless the file has a known organization or structure there seems to be just one solution to the search procedure which is to search through the records sequentially. If the file has a known organization, then more efficient search techniques are possible. For example, if the file is ordered (by the search key), then a very simple and relatively efficient search tecllniqrrc is the binary search, which successively subdivides the file into halves, and then continues the scarc11 on the half that would contain the target record. In contrast, if the fiIe has a binary tree orga.nization then the tree search technique can be used. The search problem considered in this paper ~SSIIIIIOS that the file is ordered (by key). Problems of the latter type are not discussed.

A weighted tree model for algorithms with Boolean vector operands

Abstract: Algorithms for fault tolerance decisions e.g., reconfiguration, deciding that a system is faulty, etc., frequently use operations with operands which are Boolean vectors. The bits in these vectors may represent, for example that a process or processor is up or down. Four common types of operations present in these algorithms are AND, OR, BMV (Boolean majority voting), and MASK. The authors show that by using the concept of weight of an operand, one can evaluate the structure of this type of algorithm with respect to how trustworthy their decisions are or evaluate the effect of a given input operand. >