This technical note sought to examine the ability of near-infrared reflectance spectroscopy (NIRS) to predict the chemical content and organic matter digestibility (OMD) of whole plants and the ...morphological components of forage sunflower. Empirical models for the prediction of OMD values from chemical components were developed, and their predictive ability vs. NIRS models was assessed. The total set of samples (n=147) was composed of whole plants (n=14) and morphological components (n=133) from different experiments performed at Galicia (Spain) and were scanned using a Foss NIR System 6500 instrument. The reference values of OMD corresponded to in vitro determinations (n=112 samples) from laboratory incubation tests using rumen fluid. The predictive capacity of the NIRS models was assessed by the coefficient of determination value in external validation (r2 ), showing good to excellent quality prediction of OMD and chemical components with values of r2 ≥0.88. However, the estimation of lignin did not show predictive utility (r2 =0.40). Using the NIRS models to predict the OMD of whole plants and morphological components of forage sunflower led to a decrease in the standard error in external validation, in contrast to the best empirical equation through the chemical components of samples (from ±8.25 to ±3.23%). This technical note showed that NIRS is a suitable technology, providing a rapid assessment of forage sunflower. However, these results should be considered preliminary, as they are based on a limited number of samples, and it is desirable to improve the performance of NIRS equations by increasing the dataset in future works.
The stochastic gradient descendant algorithm is one of the most popular neural network training algorithms. Many authors have contributed to modifying or adapting its shape and parametrizations in ...order to improve its performance. In this paper, the authors propose two modifications on this algorithm that can result in a better performance without increasing significantly the computational and time resources needed. The first one is a dynamic learning ratio depending on the network layer where it is applied, and the second one is a dynamic drop-out that decreases through the epochs of training. These techniques have been tested against different benchmark function to see their effect on the learning process. The obtained results show that the application of these techniques improves the performance of the learning of the neural network, especially when they are used together.
