WIP Linear parametric solution

[Affie]

Draft PR to track changes with master and discuss code
TODO

• Initialize [Feature] Parametric tree initialization #541
  • Additional states and messages will be required
• Solve
  • Calculate covariances Calculate covariances for parametric tree solve #539
  • Optim - linear works
  • NLSolve

basic usage

fg = LightDFG{SolverParams}( params=SolverParams(algorithms=[:default, :parametric]))
#add linear variables and factors ...
# Or generate a sample fg with lineStepFG
# fg = lineStepFG(7, poseEvery=1, landmarkEvery=3, posePriorsAt=collect(0:7), sightDistance=2)
#NOTE experimental and not exported. 
IIF = IncrementalInference
# bulk solve factor graph
d,st = IIF.solveFactorGraphParametric(fg)

# DEV NOTE
# For development on the tree solve (not currently working):
# build the tree
tree = wipeBuildNewTree!(fg)
# init the messages
IIF.initTreeMessageChannels!(tree)
# solve the tree
IIF.solveTreeParametric!(fg, tree)

[codecov-io]

Codecov Report

❗ No coverage uploaded for pull request base (master@ce4aa72). Click here to learn what that means.
The diff coverage is 43.47%.

@@            Coverage Diff            @@
##             master     #468   +/-   ##
=========================================
  Coverage          ?   51.73%           
=========================================
  Files             ?       35           
  Lines             ?     4413           
  Branches          ?        0           
=========================================
  Hits              ?     2283           
  Misses            ?     2130           
  Partials          ?        0

Impacted Files	Coverage Δ
src/CanonicalGraphExamples.jl	100% <ø> (ø)
src/FactorGraphTypes.jl	90.19% <ø> (ø)
src/ParametricUtils.jl	100% <ø> (ø)
src/IncrementalInference.jl	66.66% <ø> (ø)
src/SubGraphFunctions.jl	80.95% <ø> (ø)
src/DefaultNodeTypes.jl	65.21% <ø> (ø)
src/CliqStateMachineUtils.jl	29.81% <ø> (ø)
src/FGOSUtils.jl	22.91% <0%> (ø)
src/SolveTree01.jl	59.96% <0%> (ø)
src/SolverAPI.jl	45.45% <100%> (ø)
... and 6 more

---

Continue to review full report at Codecov.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update ce4aa72...1b117e7. Read the comment docs.

[dehann]

haha, sorry -- I removed the label based on PR being a 'Draft', but probably safer to have the label :-)

[Affie]

haha, sorry -- I removed the label based on PR being a 'Draft', but probably safer to have the label :-)

I was jut a bit worried that you wanted to merge it. It is keeping up to date with master so I think we can keep it separate for now. We should decide when to merge, I would like the tree to first work, but we need co-variances.

[dehann]

Swapping the tree, new CSM, and the inference in one big branch, sjoe okay! You going to have to help with the consolidation then too please :-)

[Affie]

Swapping the tree, new CSM, and the inference in one big branch, sjoe okay!

Jip, sorry. The tree foundation is in master, but I opened a can of worms with the clique index. So I first wanted to experiment and test a few ideas in the simpler parametric CSM.
I'm working towards reusing the tree. So we will need a deleteClique function.

We can merge sooner if you feel its needed. But we will end up with not fully functional code in a tag. This branch has not diverged that far from master yet, but its starting to get harder to keep it consolidated.

[dehann]

targeting slightly premature merge just after IIF v0.8.4 is tagged, coming up soon.