Some refactoring and new test #1349
Conversation
Working on this, #1343 is the WIP PR. I'm having a bit of trouble initializing the updated |
| } | ||
|
|
||
| /********************************************************************************* | ||
| // Create a hybrid nonlinear factor graph f(x0, x1, m0; z0, z1) |
There was a problem hiding this comment.
This docstring should say "Create a hybrid linear/gaussian factor graph", right?
| return discreteElimination(factors, frontalKeys); | ||
| } | ||
|
|
||
| } else { |
There was a problem hiding this comment.
Any particular reason for this nesting? Even an else if will only evaluate if discrete_only is false.
| @@ -99,6 +99,7 @@ std::function<double(const Assignment<Key> &, double)> prunerFunc( | |||
| } | |||
|
|
|||
| /* ************************************************************************* */ | |||
| // TODO(dellaert): what is this non-const method used for? Abolish it? | |||
There was a problem hiding this comment.
This is used internally to update the BayesNet with the pruned conditionals after pruning them based on the prunedDecisionTree.
@varunagrawal I refactored the test (and a bit of the elimination code). I expected a normalized DiscreteConditional on m after elimination, rather than an unnormalized factor. I propose you approve and merge this PR and then let's discuss about why that's not true.