I am trying to estimate various dynamical systems using neural network models. I am very interested in using the integrators provided by `DifferentialEquations.jl`

to do this, and the docs for that package even have nice example on parameter estimation . If possible, I would like to use DifferentialEquations with `Flux.jl`

, but so far, I cannot seem to get any example working. Here is the code I am trying

```
using DifferentialEquations
using Flux
using Flux.Tracker
b = param(-1.0)
f(t, x) = b.*x
u0 = param(1.0)
prob = ODEProblem(f, u0, (0, 1.0))
sol = solve(prob)
# output:
# ERROR: MethodError: Cannot `convert` an object of type Array{Float64,0} to an object of type TrackedArray{â€¦,Array{Float64,0}}
# This may have arisen from a call to the constructor TrackedArray{â€¦,Array{Float64,0}}(...),
# since type constructors fall back to convert methods.
```

Ultimately, I want to compare to compute `l = loss(sol.u, u_truth)`

and call `back!(l)`

to compute the gradients with respect to the parameters using `Flux`

. Is this possible in principal?