]> Git — Sourcephile - gargantext.git/blob - src/Gargantext/Core/Viz/Graph/Bridgeness.hs
[MERGE]
[gargantext.git] / src / Gargantext / Core / Viz / Graph / Bridgeness.hs
1 {-|
2 Module : Gargantext.Core.Viz.Graph.Bridgeness
3 Description : Bridgeness filter
4 Copyright : (c) CNRS, 2017-Present
5 License : AGPL + CECILL v3
6 Maintainer : team@gargantext.org
7 Stability : experimental
8 Portability : POSIX
9
10 Let be a graph Bridgeness filters inter-communities links in two ways.
11 If the partitions are known, filtering is uniform to expose the communities clearly for the beginners.
12 But
13
14
15 uniformly
16 filters inter-communities links.
17
18 TODO use Map LouvainNodeId (Map LouvainNodeId)
19 -}
20
21 {-# LANGUAGE BangPatterns #-}
22
23 module Gargantext.Core.Viz.Graph.Bridgeness -- (bridgeness)
24 where
25
26 import Data.Map.Strict (Map, fromListWith, lookup, toList, mapWithKey, elems)
27 import Data.Maybe (catMaybes)
28 import Data.Ord (Down(..))
29 import Data.Set (Set)
30 import Data.Tuple.Extra (swap)
31 import Debug.Trace (trace)
32 import Gargantext.Core.Methods.Similarities (Similarity(..))
33 import Gargantext.Prelude
34 import Prelude (pi)
35 import Graph.Types (ClusterNode(..))
36 import qualified Data.List as List
37 import qualified Data.Map.Strict as Map
38 import qualified Data.Set as Set
39 import qualified Data.Tuple.Extra as Tuple
40 import qualified Data.IntMap as Dico
41 ----------------------------------------------------------------------
42
43 type Partitions = Map (Int, Int) Double -> IO [ClusterNode]
44 type Partitions' = Map (Int, Int) Double -> IO [Set NodeId]
45 ----------------------------------------------------------------------
46 nodeId2comId :: ClusterNode -> (NodeId, CommunityId)
47 nodeId2comId (ClusterNode i1 i2) = (i1, i2)
48
49 type NodeId = Int
50 type CommunityId = Int
51
52 ----------------------------------------------------------------------
53 -- recursiveClustering : to get more granularity of a given clustering
54 -- tested with spinglass clustering only (WIP)
55 recursiveClustering' :: Partitions' -> Map (Int, Int) Double -> IO [[Set NodeId]]
56 recursiveClustering' f mp = do
57 let
58 n :: Double
59 n = fromIntegral $ Set.size
60 $ Set.unions $ List.concat
61 $ map (\(k1,k2) -> map Set.singleton [k1, k2])
62 $ Map.keys mp
63
64 t :: Int
65 t = round $ 2 * n / sqrt n
66
67 ss <- f mp
68 mapM (\s -> if Set.size s > t then f (removeNodes s mp) else pure [s]) ss
69
70 ----------------------------------------------------------------------
71 recursiveClustering :: Partitions -> Map (Int, Int) Double -> IO [ClusterNode]
72 recursiveClustering f mp = do
73 let
74 n :: Double
75 n = fromIntegral $ Set.size
76 $ Set.unions $ List.concat
77 $ map (\(k1,k2) -> map Set.singleton [k1, k2])
78 $ Map.keys mp
79
80 t :: Int
81 t = round $ 2 * n / sqrt n
82
83 (toSplit,others) <- List.span (\a -> Set.size a > t) <$> clusterNodes2sets <$> f mp
84 cls' <- mapM f $ map (\s -> removeNodes s mp) toSplit
85 pure $ setNodes2clusterNodes $ others <> (List.concat $ map clusterNodes2sets cls')
86
87
88 ----------------------------------------------------------------------
89 setNodes2clusterNodes :: [Set NodeId] -> [ClusterNode]
90 setNodes2clusterNodes ns = List.concat $ map (\(n,ns') -> toCluster n ns') $ zip [1..] ns
91 where
92 toCluster :: CommunityId -> Set NodeId -> [ClusterNode]
93 toCluster cId setNodeId = map (\n -> ClusterNode n cId) (Set.toList setNodeId)
94
95 clusterNodes2map :: [ClusterNode] -> Map NodeId Int
96 clusterNodes2map = Map.fromList . map (\(ClusterNode nId cId) -> (nId, cId))
97
98 removeNodes :: Set NodeId
99 -> Map (NodeId, NodeId) Double
100 -> Map (NodeId, NodeId) Double
101 removeNodes s = Map.filterWithKey (\(n1,n2) _v -> Set.member n1 s && Set.member n2 s)
102
103
104 clusterNodes2sets :: [ClusterNode] -> [Set NodeId]
105 clusterNodes2sets = Dico.elems
106 . Dico.fromListWith (<>)
107 . (map ((Tuple.second Set.singleton) . swap . nodeId2comId))
108
109 ----------------------------------------------------------------------
110 data Bridgeness = Bridgeness_Basic { bridgeness_partitions :: [ClusterNode]
111 , bridgeness_filter :: Double
112 }
113 | Bridgeness_Advanced { bridgeness_similarity :: Similarity
114 , bridgness_confluence :: Confluence
115 }
116 | Bridgeness_Recursive { br_partitions :: [[Set NodeId]]
117 , br_filter :: Double
118 , br_similarity :: Similarity
119 }
120
121
122 type Confluence = Map (NodeId, NodeId) Double
123
124 -- Filter Links between the Clusters
125 -- Links: Map (NodeId, NodeId) Double
126 -- List of Clusters: [Set NodeId]
127 bridgeness :: Bridgeness
128 -> Map (NodeId, NodeId) Double
129 -> Map (NodeId, NodeId) Double
130 bridgeness (Bridgeness_Recursive sn f sim) m =
131 Map.unions $ [linksBetween] <> map (\s -> bridgeness (Bridgeness_Basic (setNodes2clusterNodes s) (if sim == Conditional then pi*f else f)) m') sn
132 where
133 (linksBetween, m') = Map.partitionWithKey (\(n1,n2) _v -> Map.lookup n1 mapNodeIdClusterId
134 /= Map.lookup n2 mapNodeIdClusterId
135 ) $ bridgeness (Bridgeness_Basic clusters f) m
136 clusters = setNodes2clusterNodes (map Set.unions sn)
137 mapNodeIdClusterId = clusterNodes2map clusters
138
139
140 bridgeness (Bridgeness_Advanced sim c) m = Map.fromList
141 $ List.filter (\x -> if sim == Conditional then snd x > 0.2 else snd x > 0.02)
142 $ map (\(ks, (v1,_v2)) -> (ks,v1))
143 -- $ List.take (if sim == Conditional then 2*n else 3*n)
144 -- $ List.sortOn (Down . (snd . snd))
145 $ Map.toList
146 -- $ trace ("bridgeness3 m c" <> show (m,c))
147 $ Map.intersectionWithKey
148 (\k v1 v2 -> trace ("intersectionWithKey " <> (show (k, v1, v2))) (v1, v2)) m c
149
150
151 bridgeness (Bridgeness_Basic ns b) m = Map.fromList
152 $ List.concat
153 $ Map.elems
154 $ filterComs (round b)
155 $ groupEdges (Map.fromList $ map nodeId2comId ns) m
156
157
158 groupEdges :: (Ord comId, Ord nodeId)
159 => Map nodeId comId
160 -> Map (nodeId, nodeId) value
161 -> Map (comId, comId) [((nodeId, nodeId), value)]
162 groupEdges m = fromListWith (<>)
163 . catMaybes
164 . map (\((n1,n2), d)
165 -> let
166 n1n2_m = (,) <$> lookup n1 m <*> lookup n2 m
167 n1n2_d = Just [((n1,n2),d)]
168 in (,) <$> n1n2_m <*> n1n2_d
169 )
170 . toList
171
172 filterComs :: (Ord n1, Eq n2)
173 => Int
174 -> Map (n2, n2) [(a3, n1)]
175 -> Map (n2, n2) [(a3, n1)]
176 filterComs b m = Map.filter (\n -> length n > 0) $ mapWithKey filter' m
177 where
178 filter' (c1,c2) a
179 | c1 == c2 = a
180 -- TODO use n here
181 | otherwise = take (b * 2*n) $ List.sortOn (Down . snd) a
182 where
183 n :: Int
184 n = round $ 100 * a' / t
185 a'= fromIntegral $ length a
186 t :: Double
187 t = fromIntegral $ length $ List.concat $ elems m
188
189 --------------------------------------------------------------
190 -- Utils
191 {--
192 map2intMap :: Map (Int, Int) a -> IntMap (IntMap a)
193 map2intMap m = IntMap.fromListWith (<>)
194 $ map (\((k1,k2), v) -> if k1 < k2
195 then (k1, IntMap.singleton k2 v)
196 else (k2, IntMap.singleton k1 v)
197 )
198 $ Map.toList m
199
200 look :: (Int,Int) -> IntMap (IntMap a) -> Maybe a
201 look (k1,k2) m = if k1 < k2
202 then case (IntMap.lookup k1 m) of
203 Just m' -> IntMap.lookup k2 m'
204 _ -> Nothing
205 else look (k2,k1) m
206
207
208 {-
209 Compute the median of a list
210 From: https://hackage.haskell.org/package/dsp-0.2.5.1/docs/src/Numeric.Statistics.Median.html
211 Compute the center of the list in a more lazy manner
212 and thus halves memory requirement.
213 -}
214
215 median :: (Ord a, Fractional a) => [a] -> a
216 median [] = panic "medianFast: empty list has no median"
217 median zs =
218 let recurse (x0:_) (_:[]) = x0
219 recurse (x0:x1:_) (_:_:[]) = (x0+x1)/2
220 recurse (_:xs) (_:_:ys) = recurse xs ys
221 recurse _ _ =
222 panic "median: this error cannot occur in the way 'recurse' is called"
223 in recurse zs zs
224
225 -}