Rate Coefficients for the Collisional Excitation of Molecules: Estimates from an Artificial Neural Network - Astrophysics > Instrumentation and Methods for AstrophysicsReportar como inadecuado




Rate Coefficients for the Collisional Excitation of Molecules: Estimates from an Artificial Neural Network - Astrophysics > Instrumentation and Methods for Astrophysics - Descarga este documento en PDF. Documentación en PDF para descargar gratis. Disponible también para leer online.

Abstract: An artificial neural network ANN is investigated as a tool for estimatingrate coefficients for the collisional excitation of molecules. The performanceof such a tool can be evaluated by testing it on a dataset ofcollisionally-induced transitions for which rate coefficients are alreadyknown: the network is trained on a subset of that dataset and tested on theremainder. Results obtained by this method are typically accurate to within afactor ~ 2.1 median value for transitions with low excitation rates and ~ 1.7for those with medium or high excitation rates, although 4% of the ANN outputsare discrepant by a factor of 10 more. The results suggest that ANNs will bevaluable in extrapolating a dataset of collisional rate coefficients to includehigh-lying transitions that have not yet been calculated. For the asymmetrictop molecules considered in this paper, the favored architecture is acascade-correlation network that creates 16 hidden neurons during the course oftraining, with 3 input neurons to characterize the nature of the transition andone output neuron to provide the logarithm of the rate coefficient.



Autor: David A. Neufeld JHU

Fuente: https://arxiv.org/



DESCARGAR PDF




Documentos relacionados