|
Packit |
67cb25 |
# -*- org -*-
|
|
Packit |
67cb25 |
#+CATEGORY: monte
|
|
Packit |
67cb25 |
|
|
Packit |
67cb25 |
* Add Lattice Method and Quasi-random Method to monte carlo integration
|
|
Packit |
67cb25 |
|
|
Packit |
67cb25 |
* Fix the "No-points in left/right half space" error in miser. Random
|
|
Packit |
67cb25 |
errors like that are discouraged in a library. The routine should
|
|
Packit |
67cb25 |
iterate over the dimensions choosing a point on each side of the
|
|
Packit |
67cb25 |
bisection to ensure that the error does not occur.
|
|
Packit |
67cb25 |
|
|
Packit |
67cb25 |
* VEGAS could estimate its roundoff error, caused by dividing up into
|
|
Packit |
67cb25 |
many boxes and then summing (I know this is ridiculous for any
|
|
Packit |
67cb25 |
realistic case, but it shows up on the tests when integrating a
|
|
Packit |
67cb25 |
constant!). In particular, there are some failures reported for the
|
|
Packit |
67cb25 |
VEGAS constant tests on Windows.
|
|
Packit |
67cb25 |
|
|
Packit |
67cb25 |
* VEGAS gives a negative chisq for some cases where there are wildly
|
|
Packit |
67cb25 |
inconsistent values. Calculation should be improved so that chisq>=0.
|
|
Packit |
67cb25 |
|
|
Packit |
67cb25 |
Also, when there are inconsistent values it might make sense to scale
|
|
Packit |
67cb25 |
the error by chisq, as is often done experimentally -- at least that
|
|
Packit |
67cb25 |
way the error would be increased.
|
|
Packit |
67cb25 |
|
|
Packit |
67cb25 |
* The original VEGAS allowed to user to calculate other weighted
|
|
Packit |
67cb25 |
quantities. The appropriate way to implement it in GSL would be to
|
|
Packit |
67cb25 |
provide a separate routine which returns sample points and weights so
|
|
Packit |
67cb25 |
that the user can caculate any quantity.
|
|
Packit |
67cb25 |
|