DocumentCode :
329043
Title :
Incremental learning with and without queries in binary choice problems
Author :
Kabashima, Yoshiyuki ; Shinomoto, Shigeru
Author_Institution :
Dept. of Phys., Kyoto Univ., Japan
Volume :
2
fYear :
1993
fDate :
25-29 Oct. 1993
Firstpage :
1637
Abstract :
We consider a binary choice problem represented by a conditional probability p(s|x), where input x∈[0,1] and output s∈{-1,+1}. The function p(s|x) is assumed to be monotonically increasing with respect to x. For the best prediction, the learner must fix the optimal boundary θo determined by p(s=+1|θo)=p(s=-1|θo)=1/2 without precise knowledge of the function. Learning algorithms should work incrementally for practicality, if possible. We present incremental algorithms which ensure that the learner fixes the optimal parameter θo in the limit that the number of examples t becomes infinite. We further investigate the rate of convergence in the following two typical cases. When the learner can specify the position of input x by queries in the course of learning, the mean square deviation <(θ-θo)2> decreases as O(t-1). On the other hand, if x is chosen independently from some distribution p(x), the learner cannot attain convergence as O(t-1), but rather as O(t-45/). This rate can be accelerated by an elaborate technique up to O((In t)2t-1), as t→∞.
Keywords :
convergence of numerical methods; decision theory; learning (artificial intelligence); neural nets; optimisation; probability; binary choice problems; conditional probability; convergence rate; incremental learning; mean square deviation; neural nets; optimal boundary; queries; Acceleration; Convergence; Electronic mail; Physics; Psychology; Stochastic processes;
fLanguage :
English
Publisher :
ieee
Conference_Titel :
Neural Networks, 1993. IJCNN '93-Nagoya. Proceedings of 1993 International Joint Conference on
Print_ISBN :
0-7803-1421-2
Type :
conf
DOI :
10.1109/IJCNN.1993.716965
Filename :
716965
Link To Document :
بازگشت