Optimal generalization in perceptions
作者： O Kinouchi , N Caticha
DOI: 10.1088/0305-4470/25/23/020
关键词: Artificial neural network 、 Algorithm 、 Iterated function 、 Upper and lower bounds 、 Generalization 、 Function (mathematics) 、 Mathematics 、 Stability (learning theory) 、 Perceptron 、 Weight function
摘要: A new learning algorithm for the one-layer perceptron is presented. It aims to maximize generalization gain per example. Analytical results are obtained case of single presentation each The weight attached a Hebbian term function expected stability example in teacher perceptron. This leads obtention upper bounds ability. scheme can be iterated and numerical simulations show that it converges, within errors, theoretical optimal ability Bayes algorithm. an with maximized strategy selection examples proved that, as expected, orthogonal optimal. Exponential decay error selected examples.
harvard.edu
sci-hub.se 参考文章(11)
Ben S. Wittner, John J. Hopfield, Sara A. Solla, Lawrence D. Jackel, John S. Denker, Daniel B. Schwartz, Richard E. Howard, Large Automatic Learning, Rule Extraction, and Generalization. Complex Systems. ,vol. 1, ,(1987)
F. Vallet, J-G. Cailton, Recognition rates of the Hebb rule for learning Boolean functions Physical Review A. ,vol. 41, pp. 3059- 3065 ,(1990) , 10.1103/PHYSREVA.41.3059
F Vallet, The Hebb Rule for Learning Linearly Separable Boolean Functions: Learning and Generalization EPL. ,vol. 8, pp. 747- 751 ,(1989) , 10.1209/0295-5075/8/8/007
T L H Watkin, A Rau, Selecting examples for perceptrons Journal of Physics A. ,vol. 25, pp. 113- 121 ,(1992) , 10.1088/0305-4470/25/1/016
W Kinzel, Pal Rujan, Improving a Network Generalization Ability by Selecting Examples EPL. ,vol. 13, pp. 473- 477 ,(1990) , 10.1209/0295-5075/13/5/016
Osame Kinouchi, Nestor Caticha, Biased learning in Boolean perceptrons Physica A-statistical Mechanics and Its Applications. ,vol. 185, pp. 411- 416 ,(1992) , 10.1016/0378-4371(92)90482-6
Manfred Opper, David Haussler, Generalization performance of Bayes optimal classification algorithm for learning a perceptron. Physical Review Letters. ,vol. 66, pp. 2677- 2680 ,(1991) , 10.1103/PHYSREVLETT.66.2677
M Opper, W Kinzel, J Kleinz, R Nehl, On the ability of the optimal perceptron to generalise Journal of Physics A. ,vol. 23, ,(1990) , 10.1088/0305-4470/23/11/012
H. S. Seung, H. Sompolinsky, N. Tishby, Statistical mechanics of learning from examples Physical Review A. ,vol. 45, pp. 6056- 6091 ,(1992) , 10.1103/PHYSREVA.45.6056
L F Abbott, Learning in neural network memories Network: Computation In Neural Systems. ,vol. 1, pp. 105- 122 ,(1990) , 10.1088/0954-898X_1_1_008