Training with Noise Addition in Neural Network Solution of Inverse Problem Based on Integration of Optical Spectroscopic Methods

Abstract

In the preceding study it has been suggested to use an ensemble (integration) of optical spectroscopy methods to increase the accuracy of the solution obtained by machine learning methods. We consider joint use of Raman spectroscopy and optical absorption spectroscopy methods to determine the concentrations of heavy metal ions in water. This complex IP is solved by artificial neural networks as a machine learning method. It is demonstrated that when one of the methods is strong by its results (absorption spectroscopy), and the other one is weak (Raman spectroscopy), their joint application does not allow one to improve the results of the strong method. Moreover, the quality of the solution quickly degrades with increasing noise level in the data and with the same noise in the integrated data the patterns mentioned above remain. In practice, there may be a situation where the type and level of noise in integrated spectroscopic methods can be different. Also, it has been suggested to use noise addition when training neural networks to increase the resilience of the solution to noise in data. It has been shown that the highest resilience is achieved when the type, statistic, and noise level when training the model match the type, statistic, and noise level when the model is applied (on the test dataset).
In this study, we investigate the application of this approach to data integration of spectroscopic methods, provided that the types and noise levels in the data being integrated are different.

Speaker

Artem A. Guskov
Faculty of Physics, Moscow State University, Moscow, Russia
Russia

Discussion

Ask question