THE COMPLEXITY OF INFORMATION EXTRACTION

被引:10
作者
ABUMOSTAFA, YS [1 ]
机构
[1] CALTECH, DEPT COMP SCI, PASADENA, CA 91125 USA
关键词
COMPUTER METATHEORY - Boolean Functions - DECISION THEORY AND ANALYSIS - PROBABILITY - Random Processes;
D O I
10.1109/TIT.1986.1057209
中图分类号
TP [自动化技术、计算机技术];
学科分类号
0812 ;
摘要
In order to determine the difficulty of decision problems based on natural data, decision problems are characterized by four measures defined on a Boolean function f of N variables: the implementation cost C(f), the randomness R(f), the deterministic entropy H(f), and the complexity K(f). It is shown that: 1) C(f) approximately equals R(f) approximately equals H(f) approximately equals K(f), all measured in bits; 2) decision problems based on natural data are partially random (in the Kolmogorov sense) and have low entropy with respect to their dimensionality, and the relations between the four measures translate to lower and upper bounds on the cost of solving these problems; and 3) allowing small errors in the implementation of f saves a lot in the low-entropy case but saves nothing in the high-entropy case. If f is partially structured, the implementation cost is reduced substantially.
引用
收藏
页码:513 / 525
页数:13
相关论文
共 28 条
[1]  
Abelson H., 1978, Theoretical Computer Science, V6, P41, DOI 10.1016/0304-3975(78)90004-X
[2]   COMPOSITIONAL COMPLEXITY OF BOOLEAN FUNCTIONS [J].
ABELSON, H ;
EHRENFEUCHT, A ;
FICKETT, J ;
MYCIELSKI, J .
DISCRETE APPLIED MATHEMATICS, 1982, 4 (01) :1-10
[3]  
ABUMOSTAFA Y, 1983, THESIS CALTECH PASAD
[4]  
ABUMOSTAFA Y, 1985, FUNDAMENTAL PROBLEMS
[5]  
Aho A. V., 1974, DESIGN ANAL COMPUTER, V1st
[6]   RELATING TIME AND SPACE TO SIZE AND DEPTH [J].
BORODIN, A .
SIAM JOURNAL ON COMPUTING, 1977, 6 (04) :733-744
[7]   INFORMATION-THEORETIC COMPUTATIONAL COMPLEXITY [J].
CHAITIN, GJ .
IEEE TRANSACTIONS ON INFORMATION THEORY, 1974, 20 (01) :10-15
[8]   THEORY OF PROGRAM SIZE FORMALLY IDENTICAL TO INFORMATION-THEORY [J].
CHAITIN, GJ .
JOURNAL OF THE ACM, 1975, 22 (03) :329-340
[9]  
Duda R. O., 1973, PATTERN CLASSIFICATI
[10]  
GALLAGER RG, 1968, INFORMATION THEORY R