projects / epclust.git / blame
| Commit | Line | Data |
|---|---|---|
| 572d139a BA |
1 | simulateur : complètement aléatoire ? from Irish dataset ? |
| 2 | wavelets methods in statistics with R - p180 | |
| 3 | ||
| 4 | 00-convertir2009 et 2010.R | |
| 5 | 01-extractFeatures.R pour 2009 [utilise Stbr.R] (car on prédit 2010, pas besoin) | |
| 6 | 02-cluster2009.R | |
| 7 | 03-compute-sum-of-classes2009.R | |
| 8 | 05-cluster2WER-2009.R | |
| 9 | 06-prediction.R | |
| 10 | ||
| 11 | Essayer distance wdist du package biwavelet ? | |
| 12 | ||
| 13 | geometric structure of high dim data and dim reduction 2011 | |
| 1568db29 BA |
14 | |
| 15 | https://docs.docker.com/engine/getstarted/step_one/ | |
| 7709d507 BA |
16 | |
| 17 | A faire: | |
| 5edda192 | 18 | - finir les experiences (sur nb de classes, nb de curves / chunk, nb de procs) |
| 7709d507 | 19 | et sur d'autres architectures |
| 14cb6cf8 | 20 | |
| 14cb6cf8 BA |
21 | dans old_C_code/build : |
| 22 | cmake ../stage1/src | |
| 23 | make | |
| 24 | ||
| 25 | dans data/, lancer R puis : | |
| 26 | source("../old_C_code/wrapper.R") | |
| 65bd7506 BA |
27 | serialize("../old_C_code/build", "2009.csv","2009.bin",1) |
| 28 | library(parallel) | |
| 29 | np = detectCores() | |
| 30 | nbSeriesPerChunk = 3000 | |
| 31 | nbClusters = 20 | |
| 32 | ppam_exe("../old_C_code/build",np,"2009.bin",nbSeriesPerChunk,nbClusters) | |
| 33 | C = getMedoids("../old_C_code/build", "ppamResult.xml", "ppamFinalSeries.bin") | |
| 34 | first100series = deserialize("../old_C_code/build", "2009.bin", "2009.csv.part", "1-100") | |
| 35 | distor = getDistor("../old_C_code/build", "ppamResult.xml", "2009.bin") | |
| 5edda192 BA |
36 | |
| 37 | - interface matrice -> binaire | |
| 38 | OK | |
| 39 | ||
| 7709d507 | 40 | - courbe synchrone |
| 5edda192 | 41 | ?? |
| 7709d507 | 42 | |
| 5edda192 | 43 | Piste à explorer pour les comparaisons: H20 |
| e2a7484d BA |
44 | |
| 45 | renvoyer nombre d'individues par classe ? (+ somme ?) | |
| 46 | hypothèse : données déjà ordonnées 48 1/2H sur 365j | |
| 47 | utiliser du mixmod avec modèles allongés | |
| 48 | doit toutner sur machine plutôt standard, utilisateur "lambda" | |
| 49 | utiliser Rcpp ? | |
| 50 | ||
| dc1aa85a BA |
51 | ===== |
| 52 | ||
| 53 | trategies for upscaling | |
| 54 | From 25K to 25M : in 1000 chunks of 25K | |
| 55 | Reference values : | |
| 56 | K 0 = 200 super consumers (SC) | |
| 57 | K ∗ = 15 nal clusters | |
| 58 | 1st strategy | |
| 59 | Do 1000 times ONLY Energycon's 1st-step strategy on 25K clients | |
| 60 | With the 1000 × K 0 SC perform a 2-step run leading to K ∗ clusters | |
| 61 | ||
| 62 | --> il faut s'arranger pour que | |
| 63 | ||
| 64 | 2nd strategy | |
| 65 | Do 1000 times Energycon's 2-step strategy on 25K clients leading to | |
| 66 | 1000 × K ∗ intermediate clusters | |
| 67 | Treat the intermediate clusters as individual curves and perform a | |
| 68 | single 2-step run to get K ∗ nal clusters |