Approximation of a marine ecosystem model by artificial neural networks designed using a genetic algorithm

<p>Data from the Paper:&nbsp;Approximation of a marine ecosystem model by artificial neural&nbsp;networks designed using a genetic algorithm.</p> <p>Abstract:&nbsp;</p> <p>Marine ecosystem models are important to identify the&nbsp; processes&nbsp;that affects for example the global carbon cycle. Computation of an annually periodic solution (i.e., a steady annual cycle) for these models requires a high computational effort. To reduce&nbsp;this effort, we approximated an exemplary marine ecosystem&nbsp;model by different artificial neural networks. We used a fully connected network, then applied the sparse evolutionary training&nbsp; (SET) procedure, and finally applied a genetic algorithm (GA)&nbsp;to optimize both the &nbsp; network topology. With all three approaches, a direct approximation of the&nbsp;&nbsp;steady annual cycle&nbsp; was not sufficiently accurate. However, using the mass-corrected prediction of the ANN&nbsp;as initial concentration for additional model runs, the results were in very good agreement. &nbsp; In this way, we achieved a runtime reduction by about 15 \%. The result from the SET algorithm were comparable to those of the full network. Further application of the GA may lead to an even higher reduction.</p> <p>Content:</p> <p>Database sqlite&nbsp;<a href="https://zenodo.org/api/files/669d208b-7304-4d31-a6cc-3ad78e3544e9/ANN_Database.db">ANN_Database.db</a></p> <p>zip-files with data:&nbsp;</p> <p><a href="https://zenodo.org/api/files/669d208b-7304-4d31-a6cc-3ad78e3544e9/ANN-Data.zip">ANN-Data.zip</a>&nbsp;structure and weights of used networks</p> <p><a href="https://zenodo.org/api/files/669d208b-7304-4d31-a6cc-3ad78e3544e9/ANN-Results.zip">ANN-Results.zip</a>&nbsp;results obtained with networks</p> <p><a href="https://zenodo.org/api/files/669d208b-7304-4d31-a6cc-3ad78e3544e9/Reference-Results.zip">Reference-Results.zip</a>&nbsp;reference results and training data</p> <p>&nbsp;</p> <p>&nbsp;</p> <p>&nbsp;</p> <p>&nbsp;</p>