In a video discussing the merits of particle filters for localization, it was implied that there is some ambiguity about the complexity cost of particle filter implementations. Is this correct? Could someone explain this?
9$\begingroup$ It may just be because I am not familiar with the domain at all, but this question seems to me to lack context. I haven't watched the video; if your question is based on elements in the video, please include all necessary context in your question. $\endgroup$– Gilles 'SO- stop being evil'Mar 8, 2012 at 3:32
2$\begingroup$ You might try looking around for a paper describing particle filters. If that doesn't help, you might be able to frame the question better. $\endgroup$– SureshMar 8, 2012 at 3:52
$\begingroup$ en.wikipedia.org/wiki/Particle_filter $\endgroup$– KavehMar 8, 2012 at 4:03
$\begingroup$ While I think this question is ontopic here, it may be more successful on stats.SE. $\endgroup$– Raphael ♦Mar 8, 2012 at 6:16
It appears that the speaker feels that there isn't a definitive complexity analysis yet for the technique. This could be due to a couple of factors.
- The analysis is hard, and no one has figured it out yet.
- The technique has multiple different appropriate implementations, based on the problem context.
I'd bet on the second option, particularly in light of the speaker's comment that the technique may not be appropriate for higher dimension parameter spaces. Keep in mind that the video did not present an algorithm, it presented a very high-level discussion of a technique.
$\begingroup$ I think you might be right about point #2. I'm sure there are several different (algorithmic) ways to implement a particle filter each with complexity trade-offs, depending on the domain. $\endgroup$– DorkRawkMar 8, 2012 at 18:14