A new understanding of the denominator calculation _0615

1. Before calculating the denominator, it is to calculate the likelihood ratio of the observed data t_data and all other particles of the observed area of each particle to be updated. The energy of other particles is more diffuse than the past energy is compared;

2, but in fact phd-tbd words will be all lattice as the target, that is, at most each lattice to do a potential target or have particles there, so the calculation should also be the potential target near the observation data and randomly generated particles between

If the lattice is set to 30*40=120, the number of particles takes 100, so that a frame time is traversed to 120*100=12000. This time is almost two hours is enough, if replaced by the Jian Ronghui of the parameter 6,000 particles, it is almost 6000*120. At least not 6000*6000, it must be time-related. So we still have to change the program to traverse. So who else's article also pointed out, in order to further reduce the computational capacity, take 20 of the largest observation lattice as a target, such a calculation is 20*6000, will be much faster.

3, come, next to make changes, you can also first put MIMO in accordance with the previous plan, put it, in case the teacher to see it.

A new understanding of the denominator calculation _0615