Abstract
Linial, Mansour and Nisan introduced a learning algorithm of Boolean functions in AC 0 under the uniform distribution. Following their work, Bshouty, Jackson and Tamon, also Ohtsuki and Tomita extended the algorithm to one that is robust for noisy data. The noise process we are concerned here is as follows: for an example \(\langle x,f(x) \rangle (x \in \{ 0,1 \}^n,\ f(x) \in \{-1,1\})\), each bit of the vector x and the f(x) are obliged to change independently with probability η during a learning process. Before our present paper, we were on the assumption that the correct noise rate η or its upper bound is known in advance. In this paper, we eliminate the assumption. We estimate an upper bound of the noise rate by evaluating noisy power spectrum on the frequency domain by using a sampling trick. Thereby, by combining this procedure with Ohtsuki et al.’s algorithm, we obtain a quasipolynomial time learning algorithm that can cope with noise without knowing any information about noise in advance.
This work has been supported in part by Grants-in-Aid for Scientific Research Nos. 13680435 and 16300001 from the Ministry of Education, Culture, Sports, Science and Technology, Japan.
Access this chapter
Tax calculation will be finalised at checkout
Purchases are for personal use only
Preview
Unable to display preview. Download preview PDF.
References
Bshouty, N.H., Jackson, J.C., Tamon, C.: Uniform-distribution attribute noise learnability. In: COLT1999, pp. 75–80 (1999)
Bshouty, N.H., Jackson, J.C., Tamon, C.: Uniform-distribution attribute noise learnability. Information and Computation 187(2), 277–290 (2003)
Hagerup, T., Rub, C.: A guided tour of the Chernoff bounds. Information Processing Letters 33, 305–308 (1989)
Laird, P.D.: Learning from Good and Bad Data. Kluwer Academic Publishers, Dordrecht (1988)
Linial, L., Mansour, Y., Nisan, N.: Constant depth circuits, Fourier transform, and learnability. Journal of the ACM 40, 607–620 (1993)
Miyata, A., Tarui, J., Tomita, E.: Learning AC0 Boolean functions on attribute and label noise, Technical Report of IEICE, COMP2003-70, pp. 9–16 (2004)
Ohtsuki, K., Tomita, E.: An algorithm for learning a certain Boolean function from noisy data, Technical Report of the Department of Communications and Systems Engineering, UEC-CAS20-2, The University of Electro-Communications (2000)
Ohtsuki, K.: On learning a certain Boolean function from noisy data, Master’s dissertation at the Department of Communications and Systems Engineering in the Graduate School of Electro-Communications, The University of Electro- Communications (2001)
Author information
Authors and Affiliations
Editor information
Editors and Affiliations
Rights and permissions
Copyright information
© 2004 Springer-Verlag Berlin Heidelberg
About this paper
Cite this paper
Miyata, A., Tarui, J., Tomita, E. (2004). Learning Boolean Functions in AC 0 on Attribute and Classification Noise. In: Ben-David, S., Case, J., Maruoka, A. (eds) Algorithmic Learning Theory. ALT 2004. Lecture Notes in Computer Science(), vol 3244. Springer, Berlin, Heidelberg. https://doi.org/10.1007/978-3-540-30215-5_12
Download citation
DOI: https://doi.org/10.1007/978-3-540-30215-5_12
Publisher Name: Springer, Berlin, Heidelberg
Print ISBN: 978-3-540-23356-5
Online ISBN: 978-3-540-30215-5
eBook Packages: Springer Book Archive