MIF is super unstable and "works" really poorly unless you waste an unreasonable amount of time tuning the algorithm (or just got lucky). It seems that the current code in SSM is a fair implementation of what is in Ionides et al. 2006 and pomp 0.43-8.
To our knowledge the "frustrating" (to say the least) experience with MIF is due to MIF itself and not the SSM implementation. The tutorial example contains a version in pomp for comparison purpose (thanks to Aaron King for the original template).
Apparently MIF2 is in development. If someone is up to it pull requests are welcome!
MIF is super unstable and "works" really poorly unless you waste an unreasonable amount of time tuning the algorithm (or just got lucky). It seems that the current code in SSM is a fair implementation of what is in Ionides et al. 2006 and pomp 0.43-8.
To our knowledge the "frustrating" (to say the least) experience with MIF is due to MIF itself and not the SSM implementation. The tutorial example contains a version in pomp for comparison purpose (thanks to Aaron King for the original template). Apparently MIF2 is in development. If someone is up to it pull requests are welcome!