Loading [MathJax]/extensions/TeX/boldsymbol.js

 

 

 

Solving differential equations with RNNs

To gain some intuition on how we can use RNNs for time series, let us tailor the representation of the solution of a differential equation as a time series.

Consider the famous differential equation (Newton's equation of motion for damped harmonic oscillations, scaled in terms of dimensionless time)

\frac{d^2x}{dt^2}+\eta\frac{dx}{dt}+x(t)=F(t),

where \eta is a constant used in scaling time into a dimensionless variable and F(t) is an external force acting on the system. The constant \eta is a so-called damping.