Premium
Neural network studies. 4. An extended study of the aqueous solubility of organic compounds
Author(s) -
Bodor Nicholas,
Huang MingJu,
Harget Alan
Publication year - 1992
Publication title -
international journal of quantum chemistry
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 0.484
H-Index - 105
eISSN - 1097-461X
pISSN - 0020-7608
DOI - 10.1002/qua.560440874
Subject(s) - solubility , aqueous solution , chemistry , computational chemistry , organic chemistry
Abstract A study has been made of the effect of using different numbers of hidden units in a neural network modeling of the aqueous solubility of a wide range of organic compounds. A training set of 331 compounds was used and the trained neural network was tested on a prediction set of 19 compounds. Between two and five hidden units were used with varying numbers of iterations. Comparisons are made with the results obtained from our previous studies, one which used a neural network with 18 hidden units and another based on regression analysis. By using a smaller number of hidden units in this study, better performance has been obtained than either of the previous studies. © 1992 John Wiley & Sons, Inc.