Adding Noise During Training as a Method to Increase Resilience of Neural Network Solution of Inverse Problems: Test on the Data of Magnetotelluric Sounding Problem
In their previous studies, the authors proposed to use the approach associated with adding noise to the training set when training multilayer perceptron type neural networks to solve inverse problems. For a model inverse problem it was shown that this allows increasing the resilience of neural network solution to noise in the input data with different distributions and various intensity of noise. In the present study, the observed effect was confirmed on the data of the problem of magnetotelluric sounding. Also, maximum noise resilience (maximum quality of the solution) is generally achieved when the level of the noise in the training data set coincides with the level of noise during network application (in the test dataset). Thus, increasing noise resilience of a network when noise is added during its training is associated with the fundamental properties of multilayer perceptron neural networks and not with the properties of the data. So this method can be used solving other multi-parameter inverse problems.
KeywordsNeural networks Inverse problems Noise resilience Training with noise Regularization
The authors would like to thank E.A. Obornev, I.E. Obornev, and M.I. Shimelevich for providing the data on which this study has been performed.
- 1.Zhdanov, M.: Inverse Theory and Applications in Geophysics, 2nd edn. Elsevier, Amsterdam (2015)Google Scholar
- 3.Mohammad-Djafari, A. (ed.): Inverse Problems in Vision and 3D Tomography. Wiley, New York (2010)Google Scholar
- 4.Spichak, V.V. (ed.): Electromagnetic Sounding of the Earth’s Interior. Methods in Geochemistry and Geophysics, vol. 40. Elsevier, Amsterdam (2006)Google Scholar
- 5.Zhdanov, M.S.: Geophysical Electromagnetic Theory and Methods. Methods in Geochemistry and Geophysics, vol. 43. Elsevier, Amsterdam (2009)Google Scholar
- 16.Yin, S., Liu, C., Zhang, Z., Lin, Y., Wang, D., Tejedor, J., Zheng, T.F., Li, Y.: Noisy training for deep neural networks in speech recognition. Proc. EURASIP J. Audio Speech Music Process. 2015(2), 1–14 (2015)Google Scholar
- 17.Fadeev, V.V., Dolenko, S.A., Dolenko, T.A., Uvenkov, Ya.V., Filippova, E.M., Chubarov, V.V.: Laser diagnostics of complicated organic compounds and complexes by saturation fluorimetry. Quantum Electron. 27(6), 556–559 (1997)Google Scholar
- 18.Dolenko, S.A., Dolenko, T.A., Kozyreva, O.V., Persiantsev, I.G., Fadeev, V.V., Filippova, E.M.: Solution of inverse problem in nonlinear laser fluorimetry of organic compounds with the use of artificial neural networks. Pattern Recognit. Image Anal. 9(3), 510–515 (1999)Google Scholar
- 19.Gerdova, I.V., Dolenko, S.A., Dolenko, T.A., Persiantsev, I.G., Fadeev, V.V., Churina, I.V.: New opportunity solutions to inverse problems in laser spectroscopy involving artificial neural networks. Izvestiya Akademii Nauk. Ser. Fizicheskaya 66(8), 1116–1125 (2002)Google Scholar
- 21.Dolenko, S., Isaev, I., Obornev, E., Persiantsev, I., Shimelevich, M.: Study of influence of parameter grouping on the error of neural network solution of the inverse problem of electrical prospecting. Commun. Comput. Inf. Sci. 383, 81–90 (2013)Google Scholar
- 22.Dolenko, S., Guzhva, A., Obornev, E., Persiantsev, I., Shimelevich, M.: Comparison of adaptive algorithms for significant feature selection in neural network based solution of the inverse problem of electrical prospecting. In: Alippi, C. (ed.) ICANN 2009, Part II. Springer-Verlag, Heidelberg (2009)Google Scholar