About the Correctness of Neural Network Simulations

What is it about?

In order to build credibility in model simulations, there are verification and validation methods. However, these are not yet well-established for neural network modeling. This study proposes an adapted terminology and workflow to evaluate and boost the correctness of neural network models, even for cases where there is no experimental data to validate against.

Featured Image

Why is it important?

Creating and reporting scientific results in a way that they are easy to replicate and reproduce, is an indispensable aspect of good scientific practice. Additionally, the correctness of the results has to be ensured to not reproduce incorrect results. This work presents a rigorous workflow to ensure both, correctness and replicability, for the field of neural network simulations.