Fig. 9
From: Applications of machine learning for simulations of red blood cells in microfluidic devices
Prediction error is calculated as the distance of RBC centers for all simulated cells from A50b and centers predicted by a system of bases trained on A50a. For the first 1000 iterations, the average error is less than 2.3 μm, corresponding to about 2% of the channel length and 1/3 of the RBC diameter. In the first 2000 iterations, the average error does not exceed 3% of the channel length and half of the RBC size. The highest average error is at the order of the cell size