Fig. 2From: Automatic identification of scientific publications describing digital reconstructions of neural morphologyHyper-parameters accuracy and loss drawings used to select the most promising values. A Learning rate. B Mini-batch size. C Number of hidden units per hidden layer. D Number of hidden layers. E Optimization algorithm: Stochastic gradient descent (SGD) [38]; adaptive gradients methods RMSProp [39], and Adam [40]Back to article page