Neural network-based sensor calibration for micro and nanoelectronics applications

We studied multilayer neural network-enabled calibration of optical gas sensor systems. Such systems use fiberoptic converters so their properties can be nonlinear and nonmonotonic. We used real-world data to show the proposed calibration method's applicability. The neural network-based approach offers a higher quality of the calibration, and a multilayer neural network doe not need a training dataset to estimate the optical properties.