Artificial neural network (ANN) calibrations can encompass a vast amount of data. For instance, 25 years worth of harvest data, collected in a database comprising more than 50.000 samples, including variations in season, geography, grain type, spectroscopic scatter and temperature, are behind the robust and accurate predictions of ANN calibrations used with grain analysers.
In this paper, FOSS Senior Manager Lars Nørgaard, Chemometrician Martin Lagerholm and Research Scientist Mark Westerhaus explain the principles behind ANN calibration, demonstrating how the method is applied in the development of global calibration models on large databases.
A protein content prediction for whole wheat grain is used as a case study to demonstrate the efficiency of artificial neural networks. In particular, the case study shows how accurate qualitative and quantitative information can be extracted from complex spectroscopic databases.
If you experience any problems with the download please let us know in the comments and we will be sure to help you out.