Table 3 Linear damped oscillator: the discovered governing equations using deep IRK-SINDy, RK4-SINDy, and Conv-SINDy for various noise level \(\sigma\).

From: Implicit Runge-Kutta based sparse identification of governing equations in biologically motivated systems

Noise

Deep IRK-SINDy

RK4-SINDy

Conv-SINDy

\(\sigma = 0.01\)

\(\begin{array}{cc} & \dot{x}_{1}(t) = -0.103 x_{1}(t) + 2.000 x_{2}(t)\\ & \dot{x}_{2}(t) = -2.001 x_{1}(t) - 0.099 x_{2}(t) \end{array}\)

\(\begin{array}{cc} & \dot{x}_{1}(t) = -0.102 x_{1}(t) + 2.000 x_{2}(t)\\ & \dot{x}_{2}(t) = -2.001 x_{1}(t) - 0.099 x_{2}(t) \end{array}\)

\(\begin{array}{cc} & \dot{x}_{1}(t) = -0.102 x_{1}(t) + 1.999 x_{2}(t)\\ & \dot{x}_{2}(t) = -1.999 x_{1}(t) - 0.099 x_{2}(t) \end{array}\)

\(\sigma = 0.04\)

\(\begin{array}{cc} & \dot{x}_{1}(t) = -0.105 x_{1}(t) + 1.995 x_{2}(t)\\ & \dot{x}_{2}(t) = -2.009 x_{1}(t) - 0.097 x_{2}(t) \end{array}\)

\(\begin{array}{cc} & \dot{x}_{1}(t) = -0.104 x_{1}(t) + 1.995 x_{2}(t)\\ & \dot{x}_{2}(t) = -2.008 x_{1}(t) - 0.097 x_{2}(t) \end{array}\)g

\(\begin{array}{cc} & \dot{x}_{1}(t) = -0.104 x_{1}(t) + 1.993 x_{2}(t)\\ & \dot{x}_{2}(t) = -2.008 x_{1}(t) - 0.097 x_{2}(t) \end{array}\)

\(\sigma = 0.08\)

\(\begin{array}{cc} & \dot{x}_{1}(t) = -0.117 x_{1}(t) + 2.013 x_{2}(t)\\ & \dot{x}_{2}(t) = -1.972 x_{1}(t) - 0.105 x_{2}(t) \end{array}\)

\(\begin{array}{cc} & \dot{x}_{1}(t) = -0.115 x_{1}(t) + 2.013 x_{2}(t)\\ & \dot{x}_{2}(t) = -1.972 x_{1}(t) - 0.103 x_{2}(t) \end{array}\)

\(\begin{array}{cc} & \dot{x}_{1}(t) = -0.120 x_{1}(t) + 2.011 x_{2}(t)\\ & \dot{x}_{2}(t) = -1.964 x_{1}(t) - 0.102 x_{2}(t) \end{array}\)

\(\sigma = 0.16\)

\(\begin{array}{cc} & \dot{x}_{1}(t) = -0.147 x_{1}(t) + 1.991 x_{2}(t)\\ & \dot{x}_{2}(t) = -2.006 x_{1}(t) - 0.086 x_{2}(t) \end{array}\)

\(\begin{array}{cc} & \dot{x}_{1}(t) = -0.139 x_{1}(t) + 1.991 x_{2}(t)\\ & \dot{x}_{2}(t) = -2.002 x_{1}(t) \end{array}\)

\(\begin{array}{cc} & \dot{x}_{1}(t) = -0.144 x_{1}(t) + 1.988 x_{2}(t)\\ & \dot{x}_{2}(t) = -1.999 x_{1}(t) \end{array}\)