I fixed the bug that prevented the knots from being optimized.
However, the optimization is still not working properly since in the notebook it sets the knots to very small values, often matching the lower bound. I tried lowering the lower bound, e.g. when I set it to 0.01 we do get a sensical shape with slightly larger values for the first and last knot, but the uncertainty gets enormous.
I fixed the bug that prevented the knots from being optimized.
However, the optimization is still not working properly since in the notebook it sets the knots to very small values, often matching the lower bound. I tried lowering the lower bound, e.g. when I set it to 0.01 we do get a sensical shape with slightly larger values for the first and last knot, but the uncertainty gets enormous.