2

As I understand the basic idea of particle filter is to predict the state of the particle by generating $N$ different possible states. After that, each possible state is evaluated by a predict model (give the weight to each particle).

So...let say, if my particle is $(x, y, z)$. Then, because there are $M$ different possible values for $x$, $N$ possible values for $y$ and $L$ different possible values for $z$. The number of particles that I need to generate is $M\cdot N \cdot L$?

Thanks

sTertooy
  • 6,205
user10128
  • 143
  • The question lacks context. You may or may not have to generate all possible states depending on the context. You added the Monte Carlo tag; a Monte Carlo simulation typically doesn't generate all possible states. If your question is just whether the number of different possible values for $(x,y,z)$ is $MNL$ when there are $M$ different possible values for $x$, $N$ possible values for $y$ and $L$ different possible values for $z$, the answer is yes. – joriki Jul 07 '11 at 08:14
  • Thank Joriki. I can not tag "particle filter" or "condensation" so I had to tag "monte carlo" instead. If we dont generate all possible states, we may miss some important states, dont we? – user10128 Jul 07 '11 at 14:24
  • As I said, there's not enough context to say anything about that. – joriki Jul 07 '11 at 14:33

0 Answers0