A Note on the Prediction Algorithm when Orthogonal Design is used for Query Learning | Toshiyasu Matsushima Laboratory

Title	A Note on the Prediction Algorithm when Orthogonal Design is used for Query Learning (in Japanese)
Authors	Yoshifumi Ukita 、Daiki Koizumi 、Toshiyasu Matsushima 、Shigeichi Hirasawa
Released Year	2004
Format	Conference
Category	Knowledge information processing
Jounal Name
Jounal Page	SIG-FPAI-A402-12, pp.25-32
Published Year	2004
Published Month	11
Abstract （English）	We deal with a problem to predict the output of true Boolean function by learning from queries. If the posterior probability can be found about all Boolean functions, the optimal prediction would be determined from the point of view of Bayes decision theory. However, it is difficult to calculate such optimal prediction in practice especially in terms of the computational complexity. Therefore, we propose an approximate algorithm to take only Boolean functions such that their prior probabilities are relatively higher. In this paper, we shall show that such Boolean functions can be found efficiently when the orthogonal design is used for query learning as well as certain condition about the prior probability distribution of their functions is satisfied.
Note （English）	1
Manuscript
Presentation