Figure 3: Effect of training dataset size on the accuracy of the machine learning model. We trained individual instances of our machine learning model on datasets of progressively increasing size. Each of the resulting models was subsequently evaluated on a distinct test set comprising approximately 16×103 data points (videos). In this context, the model accuracy is defined as the average percentage of correctly predicted differential terms per video. Both the training and test datasets were synthetically created by varying the initial conditions of the “generating equation” via the data engine.