I have been applying Caputo fractional derivatives to SIR/SEIR epidemic models for the past 18 months. The memory effect captured by fractional order is fascinating but brings challenges: 1. Parameter identifiability: the fractional order α is highly sensitive to initial conditions 2. Numerical stability: L1 scheme works but Adams-Bashforth-Moulton is more accurate 3. Fitting COVID-19 data: fractional SEIR fits much better than classical ODE model Has anyone combined fractional calculus with stochastic elements? Or used Bayesian inference for α estimation?