Figure 4
A graph with four decreasing curves over epochs showing different loss trends.The horizontal axis is labeled “Epoch” and ranges from 0 to 50 with an interval of 10. The vertical axis is labeled “M S E Loss” and ranges from 0 to 2 with an interval of 0.2. Four curves are plotted and identified in the legend as “Linear loss”, “Stability loss”, “Reconstruction loss”, and “Total loss”. The “Linear loss” and “Total loss” curves start at higher values near epoch 0, with the total loss beginning close to about 1.9 and the linear loss slightly lower near about 1.85. Both curves decrease sharply within the first few epochs and approach near zero by around epoch 5, then remain nearly flat along the horizontal axis for the rest of the range. The “Reconstruction loss” curve starts at a lower value near approximately 0.1 at epoch 0 and decreases quickly to near zero within the first few epochs, then stays close to zero. The “Stability loss” curve begins very close to zero at epoch 0 and remains near the horizontal axis throughout, with only a slight change visible near the initial epochs. All four curves are smooth and decreasing, with rapid decline occurring between epochs 0 and 5 and a nearly flat trend from approximately epoch 5 to 50. Note: All numerical data values are approximated.

Training loss convergence of the DK-MPC model over 50 epochs. The plot illustrates the mean squared error (MSE) for individual loss components – linear dynamics loss, stability loss and reconstruction loss – as well as the total loss. All components show rapid convergence within the first few epochs, indicating stable and efficient learning

or Create an Account

Close subscription notice
Close access options