From 83ad6d54fcf4f0f5f5e0f00d6a22eaf78f9a3188 Mon Sep 17 00:00:00 2001 From: Stefano Pigozzi Date: Tue, 9 May 2023 16:00:06 +0200 Subject: [PATCH] Scrivi Louvain --- README.md | 98 ++++++++++++++++++++++++++++++++++++++-- scripts/2-louvain.cypher | 16 +++++++ 2 files changed, 111 insertions(+), 3 deletions(-) create mode 100644 scripts/2-louvain.cypher diff --git a/README.md b/README.md index 327e3ad..078600a 100644 --- a/README.md +++ b/README.md @@ -380,7 +380,7 @@ CALL gds.labelPropagation.write( Si osserva che l'algoritmo è giunto a convergenza. -#### Campionamento dei label +#### Campionamento delle community Si individuano gli identificatori dei label rimasti al termine dell'algoritmo: @@ -390,7 +390,7 @@ RETURN collect(DISTINCT k.communityLabelPropagation) ``` ``` -[129222, 107005, 129156, 105304, 105306, 105308, 105324, 114977, 105426, 128851, 105318, 105320, 105322, 105323, 105327, 120684, 105332, 105334, 126341, 105368, 129138, 118398, 109137, 105355, 105356, 124270, 105604, 105365, 114271, 105373, 126354, 113751, 105424, 109885, 105448, 111400, 105451, 105457, 105749, 105576, 105470, 105472, 105511, 105485, 106226, 105501, 105504, 114487, 105509, 105513, 109628, 105527, 105525, 105534, 110378, 105537, 105541, 116341, 118387, 128674, 105552, 105553, 105559, 105572, 112844, 122130, 105600, 111384, 105607, 105609, 109518, 105675, 108964, 116475, 105883, 114270, 113524, 112223, 105652, 120156, 105940, 105659, 119274, 105664, 105783, 105671, 105677, 105679, 113515, 105686, 119232, 105699, 112831, 105835, 105867, 105745, 105754, 117345, 105763, 105853, 105768, 105770, 105773, 105777, 105792, 105798, 118282, 118449, 105802, 105805, 113140, 117045, 115822, 105820, 128736, 122383, 105830, 105839, 126956, 114811, 105846, 105857, 118830, 105989, 119738, 120607, 106049, 105885, 105892, 105983, 105996, 108047, 105910, 111844, 105913, 105916, 105919, 105921, 113265, 110097, 105944, 105953, 105956, 105960, 105962, 111858, 106574, 105971, 105976, 106164, 106005, 106009, 106011, 108487, 106018, 106162, 106038, 106028, 106132, 106032, 106151, 106143, 106044, 106056, 106191, 121385, 112153, 115790, 106079, 114003, 106101, 111372, 108095, 111111, 106223, 106535, 106701, 106121, 106129, 106130, 106145, 106155, 106208, 114594, 112437, 107874, 116998, 106195, 107502, 106201, 106212, 106213, 106216, 106218, 129061, 111023, 106232, 106235, 109977, 106246, 108654, 106262, 106267, 106276, 106271, 119497, 106288, 106583, 113328, 119823, 106314, 106315, 106596, 106326, 106482, 112396, 116625, 116405, 106341, 117254, 106346, 115131, 106350, 106635, 113349, 108271, 108001, 106377, 106382, 106385, 106388, 116190, 106667, 106393, 106552, 113087, 127965, 106553, 106424, 106442, 106431, 106432, 108202, 106466, 106460, 114620, 106470, 106633, 111166, 123807, 106497, 113604, 110939, 115840, 118735, 106510, 113627, 106516, 106527, 106520, 120189, 107189, 109061, 113230, 106564, 106825, 106578, 111423, 106595, 106599, 106608, 106673, 116760, 106617, 106625, 114580, 106639, 106927, 106648, 106649, 124291, 106665, 106672, 126561, 106687, 106814, 106688, 125017, 106843, 106708, 118834, 106746, 106720, 111170, 106727, 106767, 106739, 112901, 106742, 108121, 119305, 128824, 121482, 106777, 121663, 106792, 107534, 106806, 112496, 106976, 106993, 106824, 106830, 106944, 123867, 108072, 106872, 106838, 106851, 107223, 128486, 106859, 106862, 106863, 107057, 106884, 106890, 106894, 106895, 122399, 119794, 106915, 106917, 106931, 106932, 106936, 106942, 106945, 107092, 106967, 106968, 106987, 106974, 106992, 107598, 108539, 107008, 107131, 107013, 110952, 107021, 107022, 107027, 107031, 107033, 107156, 107853, 108953, 107068, 107073, 107074, 107075, 107081, 109528, 120025, 107111, 108583, 110250, 107115, 109543, 107163, 107188, 118416, 107329, 107966, 107204, 112071, 108126, 120289, 107976, 107230, 107246, 107235, 109164, 108368, 107291, 107281, 114867, 127281, 107671, 120350, 109436, 107351, 113414, 107304, 107306, 107308, 107314, 122721, 107320, 109364, 114767, 113506, 107364, 107379, 117294, 107389, 107392, 124187, 107402, 107406, 112493, 107408, 107417, 113076, 107434, 107440, 107581, 107448, 116485, 115100, 107460, 107637, 107498, 118926, 107643, 107512, 107515, 107742, 107633, 107620, 107549, 107552, 107647, 107557, 128348, 108981, 107572, 107594, 107596, 109372, 107609, 107619, 107765, 107626, 123070, 107631, 110484, 107641, 107644, 107654, 107691, 107658, 107923, 119806, 107900, 107679, 111500, 110088, 107690, 113194, 126067, 107701, 107710, 107711, 110015, 107726, 109718, 108220, 108937, 109328, 116791, 107758, 124785, 107764, 107767, 108176, 116538, 118184, 107836, 107806, 107810, 107930, 108868, 108002, 108469, 107850, 112020, 107885, 107894, 121088, 107910, 107947, 116481, 107944, 124095, 107962, 107960, 109152, 119162] +[129222, 107005, 129156, 105304, 105306, 105308, 105324, 114977, 105426, 128851, 105318, 105320, 105322, 105323, 105327, 120684, 105332, 105334, 126341, 105368, 129138, 118398, 109137, 105355, 105356, 124270, 105604, 105365, 114271, 105373, 126354, 113751, 105424, 109885, 105448, 111400, 105451, 105457, 105749, 105576, 105470, 105472, 105511, 105485, 106226, 105501, 105504, 114487, 105509, 105513, 109628, 105527, 105525, 105534, 110378, 105537, 105541, 116341, 118387, 128674, 105552, 105553, 105559, 105572, 112844, 122130, 105600, 111384, 105607, 105609, 109518, 105675, 108964, 116475, 105883, 114270, 113524, 112223, 105652, 120156, 105940, 105659, 119274, 105664, 105783, 105671, 105677, 105679, 113515, 105686, 119232, 105699, 112831, 105835, 105867, 105745, 105754, 117345, 105763, 105853, 105768, 105770, 105773, 105777, 105792, 105798, 118282, 118449, 105802, 105805, 113140, 117045, 115822, 105820, 128736, 122383, 105830, 105839, 126956, 114811, 105846, 105857, 118830, 105989, 119738, 120607, 106049, 105885, 105892, 105983, 105996, 108047, 105910, 111844, 105913, 105916, 105919, 105921, 113265, 110097, 105944, 105953, 105956, 105960, 105962, 111858, 106574, 105971, 105976, 106164, 106005, 106009, 106011, 108487, 106018, 106162, 106038, 106028, 106132, 106032, 106151, 106143, 106044, 106056, 106191, 121385, 112153, 115790, 106079, 114003, 106101, 111372, 108095, 111111, 106223, 106535, 106701, 106121, 106129, 106130, 106145, 106155, 106208, 114594, 112437, 107874, 116998, 106195, 107502, 106201, 106212, 106213, 106216, 106218, 129061, 111023, 106232, 106235, 109977, 106246, 108654, 106262, 106267, 106276, 106271, 119497, 106288, 106583, 113328, 119823, 106314, 106315, 106596, 106326, 106482, 112396, 116625, 116405, 106341, 117254, 106346, 115131, 106350, 106635, 113349, 108271, 108001, 106377, 106382, 106385, 106388, 116190, 106667, 106393, 106552, 113087, 127965, 106553, 106424, 106442, 106431, 106432, 108202, 106466, 106460, 114620, 106470, 106633, 111166, 123807, 106497, 113604, 110939, 115840, 118735, 106510, 113627, 106516, 106527, 106520, 120189, 107189, 109061, 113230, 106564, 106825, 106578, 111423, 106595, 106599, 106608, 106673, 116760, 106617, 106625, 114580, 106639, 106927, 106648, 106649, 124291, 106665, 106672, 126561, 106687, 106814, 106688, 125017, 106843, 106708, 118834, 106746, 106720, 111170, 106727, 106767, 106739, 112901, 106742, 108121, 119305, 128824, 121482, 106777, 121663, 106792, 107534, 106806, 112496, 106976, 106993, 106824, 106830, 106944, 123867, 108072, 106872, 106838, 106851, 107223, 128486, 106859, 106862, 106863, 107057, 106884, 106890, 106894, 106895, 122399, 119794, 106915, 106917, 106931, 106932, 106936, 106942, 106945, 107092, 106967, 106968, 106987, 106974, 106992, 107598, 108539, 107008, 107131, 107013, 110952, 107021, 107022, 107027, 107031, 107033, 107156, 107853, 108953, 107068, 107073, 107074, 107075, 107081, 109528, 120025, 107111, 108583, 110250, 107115, 109543, 107163, 107188, 118416, 107329, 107966, 107204, 112071, 108126, 120289, 107976, 107230, 107246, 107235, 109164, 108368, 107291, 107281, 114867, 127281, 107671, 120350, 109436, 107351, 113414, 107304, 107306, 107308, 107314, 122721, 107320, 109364, 114767, 113506, 107364, 107379, 117294, 107389, 107392, 124187, 107402, 107406, 112493, 107408, 107417, 113076, 107434, 107440, 107581, 107448, 116485, 115100, 107460, 107637, 107498, 118926, 107643, 107512, 107515, 107742, 107633, 107620, 107549, 107552, 107647, 107557, 128348, 108981, 107572, 107594, 107596, 109372, 107609, 107619, 107765, 107626, 123070, 107631, 110484, 107641, 107644, 107654, 107691, 107658, 107923, 119806, 107900, 107679, 111500, 110088, 107690, 113194, 126067, 107701, 107710, 107711, 110015, 107726, 109718, 108220, 108937, 109328, 116791, 107758, 124785, 107764, 107767, 108176, 116538, 118184, 107836, 107806, 107810, 107930, 108868, 108002, 108469, 107850, 112020, 107885, 107894, 121088, 107910, 107947, 116481, 107944, 124095, 107962, 107960, 109152, 119162, ...] ``` Si campiona un label per verificarne i contenuti: @@ -430,7 +430,99 @@ RETURN k.name ["max32660", "cortex-a", "jtag", "bitbang", "msp432", "capacitive", "stm32l5xx", "stm32f072", ...] ``` -Si osserva una community molto ben definita di label relativi all'elettronica e alla programmazione embedded, che potrebbero individuare una categoria "Elettronica". +Si osserva una community molto ben definita di label relativi all'elettronica e alla programmazione embedded, che potrebbero individuare una categoria "Electronics and embedded programming". + +### 2️⃣ Louvain + +Si effettua un approccio diverso alla community detection, ovvero quello di usare l'*algoritmo di Louvain* per identificare i raggruppamenti che massimizzano la modularity del grafo. + +Si inizia stimando le risorse necessarie all'esecuzione dell'algoritmo: + +```cypher +CALL gds.louvain.write.estimate( + "kwds", + { + writeProperty: "communityLouvain" + } +) YIELD + nodeCount, + relationshipCount, + bytesMin, + bytesMax, + requiredMemory +``` + +| nodeCount | relationshipCount | bytesMin | bytesMax | requiredMemory | +|----------:|------------------:|---------:|---------:|---------------:| +| 24042 | 1075328 | 1546985 | 26744496 | "[1510 KiB ... 25 MiB]" | + +La quantità di memoria massima utilizzabile continua a essere all'interno le risorse disponibili a qualsiasi calcolatore moderno, pertanto si procede all'esecuzione: + +```cypher +CALL gds.louvain.write( + "kwds", + { + writeProperty: "communityLouvain" + } +) YIELD + preProcessingMillis, + computeMillis, + writeMillis, + postProcessingMillis, + nodePropertiesWritten, + communityCount, + ranLevels, + modularity, + modularities, + communityDistribution +``` + +| preProcessingMillis | computeMillis | writeMillis | postProcessingMillis | nodePropertiesWritten | communityCount | ranLevels | modularity | modularities | communityDistribution | +|---------------------|---------------|-------------|----------------------|-----------------------|----------------|-----------|--------------------|-------------------------------------------------------------|----------------------------------------------------------------------------------------------------| +| 0 | 6234 | 271 | 26 | 24042 | 1322 | 3 | 0.5470496088005096 | [0.434767324919813, 0.5403286363404538, 0.5470496088005096] | {p99: 688, min: 1, max: 2280, mean: 18.18608169440242, p90: 3, p50: 1, p999: 1994, p95: 4, p75: 1} | + +Si osserva come l'*algoritmo di Louvain* abbia creato poco più della metà delle community individuate dalla *Label Propagation*. + +##### Campionamento delle community + +Si effettuano alcuni campionamenti per verificare il contenuto delle community identificate. + +Si identificano le community formate: + +```cypher +MATCH (k:Keyword) +RETURN collect(DISTINCT k.communityLouvain) +``` + +```text +[2314, 9029, 23958, 23353, 22855, 17002, 23935, 24019, 23869, 23923, 17, 19, 21, 23853, 7113, 23830, 17993, 23061, 31, 33, 35, 36, 1825, 40, 13102, 45, 47, 23601, 23851, 13111, 68, 69, 78, 137, 161, 164, 170, 12058, 183, 185, 198, 939, 214, 217, 222, 226, 240, 238, 247, 250, 266, 272, 285, 22190, 320, 388, 596, 23966, 365, 372, 377, 9782, 384, 390, 399, 412, 458, 476, 481, 483, 486, 490, 505, 514, 515, 518, 7853, 533, 17096, 552, 559, 23922, 598, 623, 626, 632, 634, 657, 666, 669, 675, 6571, 684, 689, 724, 731, 875, 741, 745, 757, 769, 792, 814, 834, 842, 843, 7150, 908, 914, 925, 926, 929, 931, 945, 948, 959, 3367, 975, 980, 984, 1001, 1027, 1028, 1039, 1054, 1059, 1348, 8062, 1090, 1098, 1101, 1106, 1265, 1266, 1155, 1144, 1145, 1173, 1183, 1210, 12165, 1223, 1229, 1233, 14902, 1277, 1291, 1308, 1312, 1321, 1330, 1338, 1352, 1361, 19004, 1378, 1400, 1401, 1421, 1433, 5883, 1440, 1471, 1505, 1519, 1706, 1537, 1543, 1585, 1551, 1572, 1575, 1576, 1597, 1603, 1607, 1608, 17112, 1628, 1630, 1644, 1645, 1649, 1658, 1680, 1681, 1687, 1705, 3252, 1721, 1726, 5665, 1734, 1735, 1740, 1744, 1746, 1781, 1786, 1787, 1788, 1794, 4963, 1828, 1901, 1917, 6784, 1948, 1994, 2384, 2017, 2019, 2021, 2027, 2033, 4077, 2077, 12007, 2102, 2105, 2115, 2119, 2121, 2130, 7789, 2147, 2153, 2161, 2173, 2350, 2225, 2228, 2262, 2265, 2360, 2270, 3694, 2285, 2307, 2309, 2322, 2339, 17783, 2344, 5197, 2357, 2367, 2636, 2392, 2403, 2409, 2414, 2423, 2439, 11504, 2477, 2480, 2519, 2523, 2643, 2563, 2598, 2607, 2623, 2660, 2657, 2673, 2736, 2743, 2768, 2904, 2793, 3899, 2801, 2824, 10288, 2873, 2916, 2935, 2939, 2941, 2951, 2952, 2977, 3059, 3014, 3015, 3019, 3023, 3024, 3029, 3066, 3064, 3070, 3106, 3107, 3109, 3133, 3284, 3158, 3171, 3176, 3186, 3195, 3201, 3203, 3227, 3265, 3272, 19583, 3307, 3308, 3341, 3347, 3364, 3365, 3374, 3396, 3410, 19941, 3418, 3431, 3444, 7510, 3573, 3450, 3480, 3489, 3537, 3538, 3542, 3558, 3560, 3563, 3590, 3607, 3628, 3639, 3664, 3687, 3690, 3702, 3710, 3731, 3736, 3766, 3770, 3790, 3802, 22181, 3846, 3851, 3855, 3868, 3879, 4186, 3906, 3922, 4187, 4836, 6585, 3940, 4020, 6534, 3964, 3979, 3989, 4006, 4017, 4033, 4055, 4125, 4133, 4159, 4171, 4164, 6587, 12973, 4184, 4195, 4216, 4220, 4230, 5098, 4239, 4257, 4259, 4265, 4269, 4272, 4276, 4288, 4291, 4297, 4304, 4326, 4322, 4324, 13037, 4335, 4344, 4340, 4349, 4364, 4368, 4374, 4377, 4379, 4399, 4405, 4422, 4424, 4433, 4450, 4452, 4463, 4464, 4466, 4480, 23964, 4971, 4499, 4500, 4502, 4519, 4531, 4569, 4579, 10688, 4602, 4622, 4627, 7531, 5047, 4653, 4685, 4699, 4701, 4710, 4748, 4747, 4774, 4795, 5304, 4946, 4849, 4873, 4869, 4889, 4904, 4915, 4917, 4920, 4930, 4940, 4952, 6376, 4976, 4977, 4982, 5021, 5028, 5040, 5068, 5092, 5101, 5112, 5113, 5117, 10746, 5137, 5160, 5200, 5181, 5195, 5196, 5300, 5269, 5278, 5282, 5288, 5290, 5294, 5561, 5338, 5339, 5368, 5656, 5483, 5428, 5448, 5468, 5482, 5496, 5517, 5520, 6798, 6596, 17476, 5564, ...] +``` + +Si campiona la prima community restituita dalla query precedente: + +```cypher +MATCH (k:Keyword { communityLouvain: 2314 }) +RETURN k.name +``` + +```text +["gameboy-advance", "intel-8080", "sm83", "textureatlas", "rocketleaguestats", "minesweeper", "rpg-maker", "ducktyping", ...] +``` + +Si osserva che questo campione contiene numerose keyword relative a videogiochi, e in particolare a architetture di calcolo e strutture dati utilizzate per essi: si identificano pertanto due categorie, "Videogames :: Emulation" e "Videogames :: Data structures". + +Si campiona la seconda community restituita: + +```cypher +MATCH (k:Keyword { communityLouvain: 9029 }) +RETURN k.name +``` + +```text +["unittest", "snapshot-testing", "wycheproof", "stress", "nouns", "parameterisation", ...] +``` + +Si osserva che questo campione è quasi interamente relativo allo sviluppo ed esecuzione di test su codice Rust: si identifica pertanto una categoria "Testing". diff --git a/scripts/2-louvain.cypher b/scripts/2-louvain.cypher new file mode 100644 index 0000000..344a857 --- /dev/null +++ b/scripts/2-louvain.cypher @@ -0,0 +1,16 @@ +CALL gds.louvain.write( + "kwds", + { + writeProperty: "communityLouvain" + } +) YIELD + preProcessingMillis, + computeMillis, + writeMillis, + postProcessingMillis, + nodePropertiesWritten, + communityCount, + ranLevels, + modularity, + modularities, + communityDistribution \ No newline at end of file