Nicholas Krämer

This will be much easier after #353 because we have more than one object that inherits from `BayesFiltSmooth`.

Btw in case the info matters: lines such as `A = A @ B @ A.T` and `A = A @ B @ A.T + C` make up roughly a...

You have probably already found something like this, but in case not: https://stackoverflow.com/questions/3892218/how-to-test-with-pythons-unittest-that-a-warning-has-been-thrown

Great, I wasn't aware that this already exists! Unfortunately there is a speed difference:  That doesn't mean that it is not a good idea to squeeze...

Could be. Maybe it is because `Kronecker` doesn't implement `matmul` itself, so there is some nasty loopy thing going on somewhere else. I can't really judge.

> I have another benchmark request though! In PREDICT for example we compute $A P A^T + Q$. A and Q are of such a block-matrix type, but P is...

Just out of curiousity: Which direction are you thinking about here? ```python if np.isscalar(scalar) and scalar.dtype not in [np.float32, np.float64, np.integer]: scalar = np.asarray(scalar) ``` or ```python if np.isscalar(scalar) and...

continuous EKF component is covered in #231. This may serve as a template for the continuous UKF component, which will be added at some point in the future.

This will be much easier to solve after taking care of #349.

@nathanaelbosch we had an issue about this a while back. I cant find it anymore. What was the solution we agreed on? Making a step exactly to the endpoint if...