Changing and indexing prior factors after making graph about jaxfg HOT 5 CLOSED

Thanks for the quick reply I'll try it out

from jaxfg.

Hey!

The ordering of the stacked parameters should match the order that you passed the factors into StackedFactorGraph.make(factors); row i should be the mu parameter of the i-th PriorFactor.

So if you break down factors: List[FactorBase] = between_factors + prior_factors, prior_factors[i].mu.unit_complex_xy should be equivalent to graph.factor_stacks[1].factor.mu.unit_complex_xy[i, :].

Does that answer your question?

from jaxfg.

And @brentyi , how are variables in the solution ordered? Are the variables ordered by the order of appearance in the factor list, or in the pose list passed into StackedFactorGraph.make? Or something else? Thanks

from jaxfg.

The structure of the solution assignments will match the initial assignments (which needs to be explicitly passed into the solve() method), and the initial variable ordering of the assignments should parallel the factor ordering of the graph — it'll just respect whatever you pass into VariableAssignments.make_from_defaults(some iterable), VariableAssignments.make_from_dict(some dictionary), etc.

edit: this was actually aspirational behavior and untrue at the time, see #12 and #13

from jaxfg.

Thanks!

from jaxfg.