论文信息 - On the size of weights for McCulloch-Pitts neurons

On the size of weights for McCulloch-Pitts neurons

The maximum absolute value of integral weights suucient to represent any linearly separable Boolean function is investigated. It is shown that upper bounds exhibited by Muroga (1971) for rational weights satisfying the so-called normalized system of inequalities also hold for integral weights. This improves previously known upper bounds.

Michael Schmitt | M. Schmitt

[1] Robert O. Winder,et al. Threshold logic , 1971, IEEE Spectrum.

[2] A. S. Solodovnikov,et al. Systems of Linear Inequalities , 1979 .

[3] S. Muroga,et al. Theory of majority decision elements , 1961 .

[4] Marvin Minsky,et al. Perceptrons: An Introduction to Computational Geometry , 1969 .

[5] Johan Håstad,et al. On the Size of Weights for Threshold Gates , 1994, SIAM J. Discret. Math..

[6] M. Arbib. Brains, Machines, and Mathematics , 1987, Springer US.

[7] H. P. Williams. THEORY OF LINEAR AND INTEGER PROGRAMMING (Wiley-Interscience Series in Discrete Mathematics and Optimization) , 1989 .

[8] Translator-IEEE Expert staff. Machine Learning: A Theoretical Approach , 1992, IEEE Expert.

[9] P. Raghavan,et al. Learning in threshold networks , 1988, COLT '88.

[10] J. Williamson,et al. Determinants whose Elements are 0 and 1 , 1946 .

[11] Saburo Muroga,et al. Threshold logic and its applications , 1971 .