From biological data to oscillator models using SINDy - ScienceDirect
Table 2. Best discovered equations ( 𝑅 2 > 0.9 from Figure 6B) with data from the Goodwin oscillator in Equation 8 with no noise and 398 ppp using the SINDy-PI method with a library only containing terms already known from the model Figure 2. Classification of oscillator models in the model space of biological processes Figure 3. Investigation of data availability, equidistant sampling and added Gaussian noise for weak and strong timescale separation ε in the FHN model Figure 4. Addressing low-data and high noise with multiple trajectories Figure 5. Mitigating timescale separation with optimized sampling Figure 6. Enabling identification of a Hill-type response function with SINDy Figure 7. Identification of high-dimensional system by reducing set of state variables Figure 8. Overview of suggested step-by-step approach for application of SINDy on biological oscillator data Figure 9. Enabling model identification from experimental data on glycolytic oscillations in yeast Figure 10. Im