Title :
Effect of initial values in simple perception
Author :
Raudys, Sarunas ; Amari, Shun-Ichi
Author_Institution :
Inst. of Math. & Inf., Vilnius Univ., Lithuania
Abstract :
The initial value of the network weight vector can contain a large amount of information. One of the possibilities to use this information is to utilise a weighted combination of initial and final values of the weights. An alternative is the early stopping. To utilise an information contained in the initial weight vector Wˆ0 one needs to know the accuracy of determination of Wˆ0 and Wˆ 1, a final weight. The analytical expression of an optimal weighting factor greatly depends on the parameter estimation schema as well as on the learning-set size. Utilisation of the additional validation-set is a universal method
Keywords :
feedforward neural nets; initial value problems; learning (artificial intelligence); maximum likelihood estimation; statistical analysis; feedforward neural nets; initial value; initial weight vector; learning-set; maximum likelihood method; optimal weighting factor; parameter estimation; universal method; weight initialisation; Costs; Covariance matrix; Gaussian distribution; Gaussian noise; Maximum likelihood estimation; Parameter estimation; Predictive models; Statistical analysis; Statistics; Vectors;
Conference_Titel :
Neural Networks Proceedings, 1998. IEEE World Congress on Computational Intelligence. The 1998 IEEE International Joint Conference on
Conference_Location :
Anchorage, AK
Print_ISBN :
0-7803-4859-1
DOI :
10.1109/IJCNN.1998.686004
