Title :
On "learning term rewriting systems from entailment"
Author :
Krishna Rao, M.R.K.
Author_Institution :
Dept. of Inf. & Comput. Sci., King Fahd Univ. of Pet. & Miner., Dhahran, Saudi Arabia
Abstract :
Summary form only given. We study exact learning of term rewriting systems from entailment and refute a recent result by Arimura, Sakamoto and Arikawa about polynomial time learnability of k-variable linear tree translations (LTT (k)). It was incorrectly claimed that the length of derivations of LTT (k) is bounded by a polynomial in the size of the initial term. This claim led to their result on polynomial time learnability of LTT (k). We present a simple system in the class of 1-variable linear tree translations that has a derivation of exponential length. We also discuss why it is difficult to syntactically separate the rewriting systems defining polynomial functions and the rewriting systems defining exponential functions. We then identify a few requirements for polynomial time learnability of rewriting systems and discuss how these requirements may be achieved.
Keywords :
computational complexity; polynomials; rewriting systems; trees (mathematics); 1-variable linear tree translations; LTT; exponential function definition; exponential length derivation; k-variable linear tree translations; polynomial function definition; polynomial time learnability; term rewriting system learning; Computer science; Minerals; Petroleum; Polynomials;
Conference_Titel :
Computer Systems and Applications, 2003. Book of Abstracts. ACS/IEEE International Conference on
Conference_Location :
Tunis, Tunisia
Print_ISBN :
0-7803-7983-7
DOI :
10.1109/AICCSA.2003.1227526