Keras Network Learner node

Supplementary Figure 3: Details of the Keras network learner node
(A) View of the settings panel for the configuration of the model training, such as number of epochs and batch-sizes.
(B) View of the Learning monitor panel, showing the value of the loss for the training and validation batches along the training (red and blue respectively). This plot is used to estimate the progress of the model in real-time. Here, the training was run for 5 epochs with batch-size=8 (5 epochs x 15 batches/epochs = 75 batches for the full training) with 64 training images. The profile of the loss is classical of the training of machine-learning model. It keeps decreasing for the training set, while it also decreases for the test set until an optimal point (here around 45 batches) and later increases again as the model starts overfitting the training set.