Figure 4
From: Deep Cytometry: Deep learning with Real-time Inference in Cell Sorting and Flow Cytometry
Learning curves. The performance of the model is evaluated at different numbers of train examples. We trained the neural network using part of the train dataset and observed the cross-entropy errors after 100 epochs of training. To calculate the validation cross-entropy errors, we used all of the examples in the validation dataset. If the train dataset size is very small, the model does not generalize well, and the validation cross-entropy error becomes very high. When more examples from the train dataset are used, the model can generalize much better, and the validation error decreases until it settles. The train and validation cross-entropy errors almost plateau beyond a certain number of train examples, which is around 6700 examples in this case.
