Influences of the Input on ANN and QSPR of Homopolymers

  • Sun, Hong (Department of Chemistry, Tsinghua University) ;
  • Tang, Yingwu (Department of Chemistry, Tsinghua University) ;
  • Wu, Guoshi (Department of Chemistry, Tsinghua University)
  • Published : 2002.02.01

Abstract

An artificial neural network (ANN) was used to study the relationship between the glass transition temperature (T$_{g}$) and the structure of homopolymers. The input is very important for the ANN. In this paper, six kinds of input vectors were designed for the ANN. Of the six approaches, the best one gave the is T$_{g}$ of 251 polymers with a standard deviation of 8 K and a maximum error of 29 K. The trained ANN also predicted the T$_{g}$ of 20 polymers which are not included in the 251 polymers with a standard deviation of 7 K and a maximum error of 21 K. 21 K.

Keywords

References

  1. van D.W. Krevelen, Properties of Polymers, Third edition, Amsterdam, Elsevier, 1990
  2. J. Bicerano, Prediction of Polymer Properties, Marcel Dekker, NewYork, 1993
  3. C. W. UlmerII, D. A. Smith, and B. G. Sumpter, et al., Comput, Theor. Polym. Sci., 8(3/4), 311 (1998)
  4. B. G. Sumpterand D. W Noid, Macromol. Theory Simul., 3, 363 (1994)
  5. N. K. Ebube, G. O. Ababio, and C. M. Adeyeye, International Journal of Pharmaceutics, 196, 27 (2000)
  6. S. Haykin, Neural Networks - a Comprehensive Foundation, Macmillan College Publishing Company, NewYork, 1994
  7. J. Brandrup, Polymer Handbook, Third edition, New York, 1989
  8. B. R. Bhat, Modem Probability Theory - an Introductory Textbook, Second edition, NewYork, 1985