For the purpose of reduction of an error in case of determination of fire-dangerous properties of organic substances by means of a forecasting method on the basis of molecular descriptors and artificial neural networks, realized by the original computer program «Neyropaket KDS 1.0» it was decided to modify the available software product by implementation of computer library Deep Learn Toolbox. The provided library realizes modeling of artificial neural networks by «deep training». In this case the artificial neural network has two and more hidden layers. Carried out verification of data, based on some help data. It is established that the obtained data as a result of work of the modified program give a relative error in comparison with help data not exceeding 5 %.
deep training, molecular descriptors, fire safety, properties of substances, artificial neural networks
1. Tehnicheskiy reglament o trebovaniyah pozharnoy bezopasnosti: Feder. zakon ot 22 iyulya 2008 g. № 123-FZ (odob. Sovetom Federacii 11 iyulya 2008 g.) // Ros. gaz. 2008. № 163.
2. Korolev D.S., Kalach A.V., Kargashilov D.V. Prognozirovanie pozharoopasnyh svoystv veschestv i materialov s ispol'zovaniem deskriptorov i neyronnyh setey // Nauch.-teoretich. zhurn. «Vestnik BGTU im. V.G. Shuhova». 2015. № 4. S. 100-103.
3. Korolev D.S., Kalach A.V., Rudakov O.B. Prognozirovanie pozharoopasnyh svoystv veschestv // Bezopasnost' v tehnosfere. 2015. T. 56. № 5. S. 3-6.
4. Korolev D.S., Kalach A.V., Kargashilov D.V. Svidetel'stvo o gosudarstvennoy registracii programmy dlya EVM № 2016614070 «Neyropaket KDS 1.0»; pravoobladatel' FGBOU VO Voronezhskiy in-t GPS MChS Rossii. № 2016611455; zayav. 24.02.16; zaregistrirovano v reestre programm dlya EVM 16.04.16.
5. Cybernetic methods of drug design. I. Statement of the problem the perceptron approach / S.A. Hiller [et al.] // Comput. Biomed. Res. 1973. V. 6. № 5. P. 411-421.
6. Zupan J., Gasteiger J. Neural networks: a new method for solving chemical problems or just a passing phase? // Anal. Chim. Acta. 1991. V. 248. № 1. S. 1-30.
7. Freeman J.A., Skapura D.M. Neural networks: algorithms, applications and programming techniques. Addison-Wesley Publishing Company: Menlo Park, CA, 1991. 414 p.
8. Sayt Deep learning (glubokoe ili glubinnoe obuchenie). URL: http://www.insycom.ru/html/metodmat/dp.pdf (data obrascheniya: 20.11.2016).
9. Ngoc L.M., Yoon- Mo K. Quantitative prediction of lipase reaction in ionic liquids by QSAR using COSMO-RS molecular descriptors // Biochemical Engineering Journal. 2014. Vol. 87. № 5. Pp. 33-40.
10. Free S.M., Jr., Wilson J.W. A Mathematical Contribution to Structure-Activity Studies // J. Med. Chem. 1964. V. 7. № 4. P. 395-399.
11. Golender V.E., Rozenblit A.B. Logico-structural approach to computer-assisted drug design // Med. Chem. (Academic Press). 1980. V. 11. № 9. P. 299-337.
12. Korol'chenko A.Ya., Korol'chenko D.A. Pozharovzryvoopasnost' veschestv i materialov i sredstva ih tusheniya: sprav. v 2-h ch. 2-e izd., pererab. i dop. M.: Ass. «Pozhnauka», 2004. Ch. I. 713 s.
13. Korol'chenko A.Ya., Korol'chenko D.A. Pozharovzryvoopasnost' veschestv i materialov i sredstva ih tusheniya: sprav. v 2-h ch. 2-e izd., pererab. i dop. M.: Ass. «Pozhnauka», 2004. Ch. II. 774 s.
