This tiny PR resolves an issue where state variable assignments of exactly 0 resulted in 0 noise variance, and thus would throw appropriate numerical errors. This has not been relevant thusfar, as we've only added a noise kernel to states after the initial state. However, with the changes in #570 , including logging the initial state, this will no longer be the case.
This tiny PR resolves an issue where state variable assignments of exactly 0 resulted in 0 noise variance, and thus would throw appropriate numerical errors. This has not been relevant thusfar, as we've only added a noise kernel to states after the initial state. However, with the changes in #570 , including logging the initial state, this will no longer be the case.