Appendix 2—figure 1. Top: MSD (points) and power-law fits (dashed) for two different Brownian trajectories containing 1000 data points with diffusion coefficient 2.5 (red) and 0.025 (blue).

The Hurst exponent should be $\alpha =2H=1$ for both trajectories. Second Row: Simulation of the two Brownian trajectories with diffusion co-efficient 2.5 (red) and 0.025 (blue). Third Row: Local Hurst exponent estimates given by the DLFNN for the two different trajectories using a 90 point window. The averages of DLFNN Hurst exponent estimates are $\alpha =2H=1.110$ (red) and $\alpha =2H=1.114$ (blue). Bottom: Local Hurst exponent estimates of the $D=2.5$ track given by DLFNN and MSDs using a 90 point window. The average of DLFNN Hurst exponent estimates is $\alpha =2H=1.110$ and the average of MSD estimates is $\alpha =2H=0.937$.