solveTree! only returns tree

NEWS.md

@@ -24,6 +24,7 @@ The list below highlights major breaking changes, and please note that significa
 - Deprecating `approxConvBinary`, use `approxConvBelief` instead.
 - Removing obsolete `approxConvCircular`, use `approxConvBelief` instead.
 - `getSample` should return a single sample and no longer takes the N(number of samples) parameter.
+- `solveTree!` / `solveGraph!` now returns just one value `tree<:AbstractBayesTree`.  Previous version returned three values, `tree, smt, hist` (#1379).
 
 # Major changes in v0.24
 

examples/BayesTreeIllustration.jl

@@ -25,7 +25,7 @@ tree = buildTreeReset!(fg, drawpdf=true, show=true)
 # solve the factor graph and show solving progress on tree in src/JunctionTree.jl
 fg.solverParams.showtree = true
 fg.solverParams.drawtree = true
-tree, smt, hist = solveTree!(fg)
+tree = solveTree!(fg)
 
 
 ## building a new tree -- as per IIF.prepBatchTree(...)

examples/FixedPointIllustrationsSquare.jl

@@ -106,7 +106,7 @@ function runFullBatchIterations(;N=100, iters=50)
 
   FG = deepcopy(fg)
   for i in 1:iters
-    tree, smt, hist = solveTree!(fg)
+    tree = solveTree!(fg)
     push!(FG, deepcopy(fg))
   end
   return FG

examples/IllustrateAutoInit.jl

@@ -93,7 +93,7 @@ plotKDE(fg, [:x0, :x1, :x2, :x3])
 
 # Find global best likelihood solution (posterior belief)
 # After defining the problem, we can find the 'minimum free energy' solution
-tree, smt, hist = solveTree!(fg)
+tree = solveTree!(fg)
 
 # and look at the posterior belief, and notice which consensus modes stand out in the posterior
 plotKDE(fg, [:x0, :x1, :x2, :x3])

examples/MultiHypo2Door.jl

@@ -74,7 +74,7 @@ drawGraph(fg, show=true)
 tree = buildTreeReset!(fg, drawpdf=true, show=true)
 
 ## Solve graph
-tree, smt, hist = solveTree!(fg)
+tree = solveTree!(fg)
 
 # tree = buildTreeReset!(fg, drawpdf=true, show=true)
 

examples/MultiHypo3Door.jl

@@ -75,7 +75,7 @@ tree = buildTreeReset!(fg, drawpdf=true, show=true)
 
 
 ## Solve graph
-tree, smt, hist = solveTree!(fg)
+tree = solveTree!(fg)
 
 ## Plotting functions below
 

examples/RobotFourDoor.jl

@@ -62,7 +62,7 @@ plotKDE(fg, :x4)
 addFactor!(fg,[:x4], doorPrior)
 
 # solve over all data
-tree, smt, hists = solveTree!(fg)
+tree = solveTree!(fg)
 
 # list variables and factors in fg
 @show ls(fg) # |> sortDFG

examples/squarefixedpoint.jl

@@ -38,7 +38,7 @@ doautoinit!(fg, :xy)
 initManual!(fg, :x, randn(1,100))
 
 # find solution
-tree, smt, hist = solveTree!(fg)
+tree = solveTree!(fg)
 
 ## plot the result
 plotKDE(map(x->getKDE(fg,x), [:x; :xy]))

src/IncrementalInference.jl

@@ -14,7 +14,11 @@ using Reexport
 
 using Manifolds
 
-export ℝ, AbstractManifold, Euclidean, Circle
+export ℝ, AbstractManifold
+# common groups -- preferred defaults at this time.
+export TranslationGroup, CircleGroup
+# common non-groups -- TODO still teething problems to sort out in IIF v0.25-v0.26.
+export Euclidean, Circle
 
 import NLsolve
 import NLSolversBase
@@ -119,9 +123,9 @@ export FunctorInferenceType, PackedInferenceType
 export AbstractPrior, AbstractRelative, AbstractRelativeRoots, AbstractRelativeMinimize
 
 # not sure if this is necessary
-export convert
+export convert, *
 
-export *,
+export
   CSMHistory,
   # getTreeCliqsSolverHistories,
 

src/SolverAPI.jl

@@ -254,7 +254,7 @@ DevNotes
 Example
 ```julia
 # pass in old `tree` to enable compute recycling -- see online Documentation for more details
-tree, smt, hist = solveTree!(fg [,tree])
+tree = solveTree!(fg [,tree])
 ```
 
 Related
@@ -372,8 +372,7 @@ function solveTree!(dfgl::AbstractDFG,
   oldtree.eliminationOrder = tree.eliminationOrder
   oldtree.buildTime = tree.buildTime
 
-  hist = !opt.async ? fetchCliqHistoryAll!(smtasks) : hist
-
+
   if opt.drawtree && opt.async
     @warn "due to async=true, only keeping task pointer, not stopping the drawtreerate task!  Consider not using .async together with .drawtreerate != 0"
     push!(smtasks, treetask)
@@ -383,10 +382,11 @@ function solveTree!(dfgl::AbstractDFG,
 
   # if debugging and not async then also print the CSMHistory
   if opt.dbg && !opt.async
-    printCSMHistorySequential(hist, joinLogPath(dfgl,"HistoryCSMAll.txt") )
+    hists = !opt.async ? fetchCliqHistoryAll!(smtasks) : hist
+    printCSMHistorySequential(hists, joinLogPath(dfgl,"HistoryCSMAll.txt") )
   end
-
-  return oldtree, smtasks, hist
+  
+  return oldtree
 end
 
 """

src/TreeDebugTools.jl

@@ -688,7 +688,7 @@ fg = initfg()
 
 fsy = getTreeAllFrontalSyms(fg, tree) # for later use
 # perform inference to find the factor graph marginal posterior estimates
-tree, smt, hist = solveTree!(fg, recordcliqs=fsy)
+tree = solveTree!(fg, recordcliqs=fsy)
 
 # generate frames in standard location /tmp/caesar/csmCompound/
 #  requires: sudo apt-get install graphviz

src/services/HeatmapSampler.jl

@@ -59,6 +59,10 @@ function fitKDE(support,
   kde!(support, kernel_bw, weights)
 end
 
+
+global allthres = Dict{Int, Any}()
+
+
 function HeatmapDensityRegular( data::AbstractMatrix{<:Real}, 
                                 domain::Tuple{<:AbstractVector{<:Real},<:AbstractVector{<:Real}},
                                 level::Real,
@@ -68,8 +72,12 @@ function HeatmapDensityRegular( data::AbstractMatrix{<:Real},
                                 bw_factor::Real=0.7,  # kde spread between domain points 
                                 N::Int=10000  )
   #
+  global allthres
+
   # select the support from raw data
   support_, weights_, roi = getLevelSetSigma(data, level, sigma, domain...; sigma_scale=sigma_scale)
+  allthres[length(allthres)+1] = roi
+
   # constuct a pre-density from which to draw intermediate samples
   density_ = fitKDE(support_, weights_, domain...; bw_factor=bw_factor)
   pts_preIS, = sample(density_, N)

test/TestCSMMultihypo.jl

@@ -72,7 +72,7 @@ getSolverParams(fg).limititers = 30 # previous runaway CSM issue due to excessiv
 # getSolverParams(fg).drawtree = false
 # getSolverParams(fg).showtree = false
 
-tree, smt, hist = solveTree!(fg, recordcliqs=ls(fg))
+tree = solveTree!(fg, recordcliqs=ls(fg))
 
 
 # drawGraph(fg)

test/fourdoortest.jl

@@ -51,7 +51,7 @@ addFactor!(fg,[:x3;:x4], LinearRelative( Normal(200.0,4.0)))
 addFactor!(fg,[:x4], doorPrior)
 
 # solve over all data
-tree, smt, hists = solveTree!(fg)
+tree = solveTree!(fg)
 
 ##
 

test/priorusetest.jl

@@ -35,7 +35,7 @@ addFactor!(fg, [:x0; :x1], LinearRelative(Normal(0.0, 0.01)))
 addFactor!(fg, [:x1; :x2], LinearRelative(Normal(0.0, 0.01)))
 
 #solve
-tree, smt, hist = solveTree!(fg)
+tree = solveTree!(fg)
 x0_m = getKDEMean(getKDE(getVariable(fg, :x0)))[1]
 x1_m = getKDEMean(getKDE(getVariable(fg, :x1)))[1]
 x2_m = getKDEMean(getKDE(getVariable(fg, :x2)))[1]
@@ -88,7 +88,7 @@ addFactor!(fg, [:x2; :l0], LinearRelative(Normal(0, 0.01)))
 addFactor!(fg, [:x2; :l1], LinearRelative(Normal(0, 0.01)))
 
 #solve
-tree, smt, hist = solveTree!(fg)
+tree = solveTree!(fg)
 
 x0_m = getKDEMean(getKDE(getVariable(fg, :x0)))[1]
 x1_m = getKDEMean(getKDE(getVariable(fg, :x1)))[1]

test/testBasicCSM.jl

@@ -55,7 +55,7 @@ getSolverParams(dfg).limititers = 50
 ## getSolverParams(dfg).async = true
 
 
-tree, smtasks, hist = solveTree!(dfg) #, recordcliqs=ls(dfg))
+tree = solveTree!(dfg) #, recordcliqs=ls(dfg))
 
 pts_ = getBelief(dfg, :c) |> getPoints
 TensorCast.@cast pts[i,j] := pts_[j][i]

test/testBasicGraphs.jl

@@ -28,12 +28,12 @@ addFactor!(fg, [:x0;], Prior(Normal(0.0,1.0)))
 @test !isSolved(getVariable(fg, :x0))
 
 # run solver once
-tree, smt, hist = solveTree!(fg)
+tree = solveTree!(fg)
 
 @test getSolvedCount(fg, :x0) == 1
 @test isSolved(fg, :x0)
 
-tree, smt, hist = solveTree!(fg)
+tree = solveTree!(fg)
 
 @test getSolvedCount(fg, :x0) == 2
 @test isSolved(fg, :x0)
@@ -61,7 +61,7 @@ fg = initfg()
 addVariable!(fg, :x0, ContinuousScalar)
 addFactor!(fg, [:x0;], Prior(Normal(1000.0,1.0)))
 
-tree, smt, hist = solveTree!(fg)
+tree = solveTree!(fg)
 
 # check mean and covariance
 @test abs((getBelief(fg, :x0) |> getKDEMean)[1]-1000) < 0.5
@@ -80,7 +80,7 @@ addVariable!(fg, :x0, ContinuousScalar)
 addFactor!(fg, [:x0;], Prior(Normal(0.0,1.0)))
 addFactor!(fg, [:x0;], Prior(Normal(0.0,1.0)))
 
-tree, smt, hist = solveTree!(fg)
+tree = solveTree!(fg)
 
 # check mean and covariance
 @test (getBelief(fg, :x0) |> getKDEMean .|> abs)[1] < 0.4
@@ -101,7 +101,7 @@ addFactor!(fg, [:x0;], Prior(Normal(0.0,1.0)))
 addFactor!(fg, [:x0;], Prior(Normal(0.0,1.0)))
 addFactor!(fg, [:x0;], Prior(Normal(0.0,1.0)))
 
-tree, smt, hist = solveTree!(fg)
+tree = solveTree!(fg)
 
 # check mean and covariance
 @test (getBelief(fg, :x0) |> getKDEMean .|> abs)[1] < 0.4
@@ -122,7 +122,7 @@ addVariable!(fg, :x0, ContinuousScalar)
 addFactor!(fg, [:x0;], Prior(Normal(-1.0,1.0)))
 addFactor!(fg, [:x0;], Prior(Normal(+1.0,1.0)))
 
-tree, smt, hist = solveTree!(fg)
+tree = solveTree!(fg)
 
 # check mean and covariance -- should be zero
 @test (getBelief(fg, :x0) |> getKDEMean .|> abs)[1] < 0.8
@@ -142,7 +142,7 @@ addVariable!(fg, :x0, ContinuousScalar)
 addFactor!(fg, [:x0;], Prior(Normal(-1.0-1000,1.0)))
 addFactor!(fg, [:x0;], Prior(Normal(+1.0-1000,1.0)))
 
-tree, smt, hist = solveTree!(fg)
+tree = solveTree!(fg)
 
 # check mean and covariance -- should be zero
 @test abs((getBelief(fg, :x0) |> getKDEMean)[1] + 1000) < 0.6
@@ -165,7 +165,7 @@ addFactor!(fg, [:x0;], Prior(Normal(0.0,1.0)))
 addFactor!(fg, [:x1;], Prior(Normal(0.0,1.0)))
 addFactor!(fg, [:x0;:x1;], LinearRelative(Normal(0.0,10.0)))
 
-tree, smt, hist = solveTree!(fg)
+tree = solveTree!(fg)
 
 # check mean and covariance -- should be zero
 @test (getBelief(fg, :x0) |> getKDEMean .|> abs)[1] < 0.6
@@ -191,7 +191,7 @@ addFactor!(fg, [:x0;], Prior(Normal(-1.0,1.0)))
 addFactor!(fg, [:x1;], Prior(Normal(+1.0,1.0)))
 addFactor!(fg, [:x0;:x1;], LinearRelative(Normal(0.0,10.0)))
 
-tree, smt, hist = solveTree!(fg)
+tree = solveTree!(fg)
 
 # check mean and covariance -- should be near each prior
 @test abs((getBelief(fg, :x0) |> getKDEMean)[1]+1) < 0.75
@@ -220,7 +220,7 @@ addFactor!(fg, [:x2;], Prior(Normal(+1.0,1.0)))
 addFactor!(fg, [:x0;:x1;], LinearRelative(Normal(0.0,1.0)))
 addFactor!(fg, [:x1;:x2;], LinearRelative(Normal(0.0,1.0)))
 
-tree, smt, hist = solveTree!(fg)
+tree = solveTree!(fg)
 
 
 
@@ -264,7 +264,7 @@ addFactor!(fg, [:x3;:x4;], LinearRelative(Normal(0.0,1.0)))
 # #1196
 drawGraph(fg, filepath="testgraphplot/myfg.dot", show=false)
 
-tree, smt, hist = solveTree!(fg, storeOld=true)
+tree = solveTree!(fg, storeOld=true)
 
 # using KernelDensityEstimatePlotting
 # plotKDE((x->getBelief(fg,x)).([:x0;:x1;:x2;:x3;:x4]))
@@ -331,7 +331,7 @@ pts_ = getPoints(getBelief(fg, :x0))
 TensorCast.@cast pts[i,j] := pts_[j][i]
 X0 = pts |> deepcopy
 
-tree, smt, hist = solveTree!(fg)
+tree = solveTree!(fg)
 
 # values after solve
 pts_ = getPoints(getBelief(fg, :x0))

test/testBasicParametric.jl

@@ -84,7 +84,7 @@ foreach(x->getSolverData(getVariable(fg,x.first),:parametric).val[1] .= x.second
 # getSolverParams(fg).async = true
 getSolverParams(fg).graphinit = false
 
-tree2, smt, hist = IIF.solveTree!(fg; algorithm = :parametric) #, recordcliqs=ls(fg))
+tree2 = IIF.solveTree!(fg; algorithm = :parametric) #, recordcliqs=ls(fg))
 
 
 for i in 0:10
@@ -150,7 +150,7 @@ foreach(x->getSolverData(getVariable(fg,x.first),:parametric).val[1] .= x.second
   #   global smt
   #   global hist
 #force message passing with manual variable order
-tree2, smt, hist = solveTree!(fg; algorithm=:parametric, eliminationOrder=[:x0, :x2, :x1])
+tree2 = solveTree!(fg; algorithm=:parametric, eliminationOrder=[:x0, :x2, :x1])
 # end
 foreach(v->println(v.label, ": ", DFG.getSolverData(v, :parametric).val), getVariables(fg))
 
@@ -189,7 +189,7 @@ foreach(x->getSolverData(getVariable(fg,x.first),:parametric).val[1] .= x.second
 # fg.solverParams.drawtree = true
 # fg.solverParams.dbg = false
 getSolverParams(fg).graphinit = false
-tree2, smt, hist = IIF.solveTree!(fg; algorithm=:parametric)
+tree2 = IIF.solveTree!(fg; algorithm=:parametric)
 
 # print results
 if false