[Pipeline] clustering with C++ Louvain bindings, ok.

src/Gargantext/Pipeline.hs

@@ -22,17 +22,16 @@ import Gargantext.Core (Lang(FR))
 import Gargantext.Prelude
 
 import Gargantext.Viz.Graph.Index (score, createIndices, toIndex)
-import Gargantext.Viz.Graph.Distances.Matrice (distributional)
+import Gargantext.Viz.Graph.Distances.Matrice (conditional)
 import Gargantext.Text.Metrics.Occurrences (cooc, removeApax)
 import Gargantext.Text.Terms (TermType(Multi, Mono), extractTerms)
 import Gargantext.Text.Context (splitBy, SplitContext(Sentences))
 
-import Data.Graph.Clustering.Louvain (bestpartition)
-import Data.Graph.Clustering.Louvain.Utils (map2graph)
+import Data.Graph.Clustering.Louvain.CplusPlus (cLouvain)
 
 pipeline path = do
   -- Text  <- IO Text <- FilePath
-  text     <- readFile path  
+  text     <- readFile path
   let contexts = splitBy (Sentences 3) text
   myterms <- extractTerms Multi FR contexts
   
@@ -40,11 +39,12 @@ pipeline path = do
   -- TODO    groupBy (Stem | GroupList)
   
   let myCooc = removeApax $ cooc myterms
-  
   -- Cooc -> Matrix
-  let theScores = M.filter (/=0) $ score distributional myCooc
+  let theScores = M.take 350 $ M.filter (>0) $ score conditional myCooc
   let (ti, _) = createIndices theScores
-  
-  -- Matrix -> Clustering -> Graph -> JSON
-  pure $ bestpartition False $ map2graph $ toIndex ti theScores
+--
+----  -- Matrix -> Clustering -> Graph -> JSON
+----  pure $ bestpartition False $ map2graph $ toIndex ti theScores
+  partitions <- cLouvain $ toIndex ti theScores
+  pure partitions
 

src/Gargantext/Viz/Graph.hs

@@ -60,8 +60,3 @@ $(deriveJSON (unPrefix "g_") ''Graph)
 
 
 
-
-
-
-
-

src/Gargantext/Viz/Graph/Distances/Matrice.hs

@@ -89,8 +89,21 @@ type Matrix' a = Acc (Matrix a)
 type InclusionExclusion    = Double
 type SpecificityGenericity = Double
 
-conditional :: Matrix Double -> (Matrix InclusionExclusion, Matrix SpecificityGenericity)
-conditional m = (run $ ie (use m), run $ sg (use m))
+
+miniMax :: Matrix' Double -> Matrix' Double
+miniMax m = map (\x -> ifThenElse (x > miniMax') x 0) m
+  where
+    miniMax' = (the $ minimum $ maximum m)
+
+conditional :: Matrix Int -> Matrix Double
+conditional m = run (miniMax $ proba r $ map fromIntegral $ use m)
+  where
+    r :: Rank
+    r = rank' m
+
+
+conditional' :: Matrix Double -> (Matrix InclusionExclusion, Matrix SpecificityGenericity)
+conditional' m = (run $ ie (use m), run $ sg (use m))
   where
 
     ie :: Matrix' Double -> Matrix' Double
@@ -115,14 +128,10 @@ conditional m = (run $ ie (use m), run $ sg (use m))
 -- | Distributional Distance
 
 distributional :: Matrix Int -> Matrix Double
-distributional m = run $ filter $ ri (map fromIntegral $ use m)
+distributional m = run $ miniMax $ ri (map fromIntegral $ use m)
   where
     n    = rank' m
     
-    miniMax m = map (\x -> ifThenElse (x > miniMax') x 0) m
-      where
-        miniMax' = (the $ minimum $ maximum m)
-    
     filter  m = zipWith (\a b -> max a b) m (transpose m)
     
     ri mat = zipWith (/) mat1 mat2