Merge ~kub/dpcs:database into dpcs:master

[kub]

30e7f02... by kub on 2016-03-06

added classification

Unmerged commits

30e7f02... by kub on 2016-03-06

added classification

b293be8... by kub on 2016-03-06

er a commit message to explain why this merge is necessary,

ae72a32... by kub on 2016-03-06

some tidying

3b2fdc5... by kub on 2016-03-06

Merge branch 'database' of git+ssh://git.launchpad.net/~kub/dpcs into database

7afc1a4... by kub on 2016-03-06

simple script to create tables

33d621e... by kub on 2016-03-06

simple script to create tables

diff --git a/docs/classificationB/decision_trees_logistic_regression.odt b/docs/classificationB/decision_trees_logistic_regression.odt
 new file mode 100644
index 0000000..96144f6
 Binary files /dev/null and b/docs/classificationB/decision_trees_logistic_regression.odt differ
diff --git a/docs/classificationB/decision_trees_logistic_regression.pdf b/docs/classificationB/decision_trees_logistic_regression.pdf
 new file mode 100644
index 0000000..6f2564e
 Binary files /dev/null and b/docs/classificationB/decision_trees_logistic_regression.pdf differ
diff --git a/docs/classificationB/ml.Rmd b/docs/classificationB/ml.Rmd
 new file mode 100644
index 0000000..d7e03b8
--- /dev/null
+++ b/docs/classificationB/ml.Rmd
@@ -0,0 +1,75 @@
+---
+title: "ML"
+output: word_document
+---
+
+```{r setup, include=FALSE}
+knitr::opts_chunk$set(echo = TRUE)
+```
+
+```{r echo = FALSE, message = FALSE, results="hide"}
+library("glmnet")
+library("tree")
+library("tictoc")
+
+dataSetup <- function(catNumber, trainingExamples, noise, testExamples){
+rule <<- rbinom(catNumber, 1, 0.5)
+xTrain <<- matrix(rbinom(trainingExamples * catNumber, 1, 0.5), trainingExamples, catNumber)
+yTrain <<- apply(xTrain == rule, 1, all)
+noiseMask <<- matrix(rbinom(trainingExamples, 1, noise), trainingExamples, 1)
+yTrain <<- xor(yTrain, noiseMask)
+xTest <<- matrix(rbinom(testExamples * catNumber, 1, 0.5), testExamples, catNumber)
+yTest <<- apply(xTest == rule, 1, all)
+}
+
+logistic <- function(catNumber, trainingExamples, noise, testExamples){
+  
+  dataSetup(catNumber, trainingExamples, noise, testExamples)
+  tic()
+  regressionModel <- glmnet(xTrain, yTrain, family = "binomial")
+  regressionPrediction <- predict(regressionModel, newx=xTest, type = "class",  s = 1e-3)
+  regressionTime <- toc()
+  regressionError <- sum(!regressionPrediction == yTest) / testExamples
+  regressionTime = regressionTime$toc - regressionTime$tic
+  return(c(regressionError, regressionTime))
+}
+
+tree2 <- function(catNumber, trainingExamples, noise, testExamples){
+  
+  dataSetup(catNumber, trainingExamples, noise, testExamples)
+  dfTrain <- data.frame(yTrain, xTrain)
+  dfTest <- data.frame(yTest, xTest)
+  tic()
+  treeFit <- tree('yTrain~.', dfTrain)
+  treePrediction <- predict(treeFit, dfTest, type = "vector")
+  treePrediction <- treePrediction >= 0.5
+  treeError <- sum(!treePrediction == yTest) / testExamples
+  treeTime <- toc()
+  treeTime <- treeTime$toc - treeTime$tic
+  return(c(treeError, treeTime))
+}
+
+#parameters
+catNumber <- 5
+trainingExamples <- 1000
+noise <- 0.3
+testExamples <- 100
+series <- 100
+
+logisticPoints <- logistic(catNumber, trainingExamples, noise, testExamples)
+treePoints <- tree2(catNumber, trainingExamples, noise, testExamples)
+
+for(i in 2:series){#idiotyczne
+  logisticPoints <- cbind(logisticPoints, logistic(catNumber, trainingExamples, noise, testExamples))
+  treePoints <- cbind(treePoints, tree2(catNumber, trainingExamples, noise, testExamples))
+}
+
+
+plot(logisticPoints[1,], logisticPoints[2,], col = "red", xlim = c(0,1), ylim = c(0,0.3), xlab = "Error", ylab = "Time", main ="blue trees, red logreg")
+par(new = TRUE)
+plot(treePoints[1,], treePoints[2,], col = "blue", xlim = c(0,1), ylim = c(0,0.3), xlab = "", ylab ="")
+par(new = FALSE)
+
+
+```
+
diff --git a/docs/classificationB/ml_multinomial.Rmd b/docs/classificationB/ml_multinomial.Rmd
 new file mode 100644
index 0000000..9ddee67
--- /dev/null
+++ b/docs/classificationB/ml_multinomial.Rmd
@@ -0,0 +1,100 @@
+---
+title: "ML"
+output: word_document
+---
+
+```{r setup, include=FALSE}
+knitr::opts_chunk$set(echo = TRUE)
+```
+
+```{r echo = FALSE, message = FALSE, results="hide"}
+library("glmnet")
+library("tree")
+library("tictoc")
+
+dataSetup <- function(npred, class.size, nclass){
+
+class.labels <- as.character(1:nclass)
+class.predictors <- replicate(nclass,{
+  sample(1:npred, size=4)
+})
+colnames(class.predictors) <- class.labels
+  
+x <- matrix(rbinom(npred*nclass*class.size, 1, 0.05), nrow=(class.size*nclass), ncol = npred) #random x
+y <- rep(class.labels, class.size)
+
+#set appriopriate cells in x to 1
+for (label in class.labels){
+  x[y==label, class.predictors[,label]] = 1
+}
+half <- floor(dim(x)[1]/2)
+train <- 1:half
+test <- (half+1):(2*half)
+
+xTrain <<- x[train,]
+yTrain <<- y[train]
+xTest <<- x[test,]
+yTest <<- y[test]
+
+#rule <<- rbinom(catNumber, 1, 0.5)
+#xTrain <<- matrix(rbinom(trainingExamples * catNumber, 1, 0.5), trainingExamples, catNumber)
+#yTrain <<- apply(xTrain == rule, 1, all)
+#noiseMask <<- matrix(rbinom(trainingExamples, 1, noise), trainingExamples, 1)
+#yTrain <<- xor(yTrain, noiseMask)
+#xTest <<- matrix(rbinom(testExamples * catNumber, 1, 0.5), testExamples, catNumber)
+#yTest <<- apply(xTest == rule, 1, all)
+}
+
+logistic <- function(npred, class.size, nclass){
+  
+  dataSetup(npred, class.size, nclass)
+  tic()
+  regressionModel <- glmnet(xTrain, yTrain, family = "multinomial")
+  regressionPrediction <- predict(regressionModel, newx=xTest, type = "class", s= 1e-3) #* class.size * nclass) # 1e-8 * catNumber * trainingExamples)
+  regressionTime <- toc()
+  regressionError <- sum(!regressionPrediction == yTest) / (length(yTest))
+  regressionTime = regressionTime$toc - regressionTime$tic
+  return(c(regressionError, regressionTime))
+}
+
+tree2 <- function(npred, class.size, nclass){
+  
+  dataSetup(npred, class.size, nclass)
+  dfTrain <- data.frame(yTrain, xTrain)
+  dfTest <- data.frame(yTest, xTest)
+  tic()
+  treeFit <- tree('yTrain~.', dfTrain)
+  treePrediction <- predict(treeFit, dfTest, type = "class")
+  #treePrediction <- treePrediction >= 0.5
+  treeError <- sum(!treePrediction == yTest) / (length(yTest))
+  treeTime <- toc()
+  treeTime <- treeTime$toc - treeTime$tic
+  return(c(treeError, treeTime))
+}
+
+#parameters
+npred <- 20
+class.size <-50
+nclass <- 5
+#catNumber <- 5
+#trainingExamples <- 100
+#noise <- 0.3
+#testExamples <- 100
+series <- 20
+
+logisticPoints <- logistic(npred, class.size, nclass)
+treePoints <- tree2(npred, class.size, nclass)
+
+for(i in 2:series){#idiotyczne
+  logisticPoints <- cbind(logisticPoints, logistic(npred, class.size, nclass))
+  treePoints <- cbind(treePoints, tree2(npred, class.size, nclass))
+}
+
+
+plot(logisticPoints[1,], logisticPoints[2,], col = "red",xlim = c(0,0.2), ylim = c(0,0.4),  xlab = "Error", ylab = "Time", main ="blue trees, red logreg")
+par(new = TRUE)
+plot(treePoints[1,], treePoints[2,], col = "blue",xlim = c(0,0.2), ylim = c(0,0.4), xlab = "", ylab ="")
+par(new = FALSE)
+
+```
+
diff --git a/server/database/tables.sql b/server/database/tables.sql
 new file mode 100644
index 0000000..c137897
--- /dev/null
+++ b/server/database/tables.sql
@@ -0,0 +1,8 @@
+DROP DATABASE IF EXISTS crash_fixer;
+CREATE DATABASE crash_fixer;
+\c crash_fixer;
+CREATE TABLE applications(id INT PRIMARY KEY NOT NULL, name VARCHAR(250), version VARCHAR(250));
+CREATE TABLE system_info(id INT PRIMARY KEY, version VARCHAR(250));
+CREATE TABLE solutions(id INT PRIMARY KEY NOT NULL, details TEXT);
+CREATE TABLE crash_groups(id INT PRIMARY KEY NOT NULL, solution_id INT REFERENCES solutions(id));
+CREATE TABLE crash_reports(id INT PRIMARY KEY NOT NULL, url VARCHAR(250), crash_group_id INT REFERENCES crash_groups(id), application_id INT REFERENCES applications(id), system_info_id INT REFERENCES system_info(id), exit_code INT, stderr_output TEXT);