tests and docs (goldens and updating user guide)

R/tests/testdir_golden/runit_glm2_11_golden.R

@@ -0,0 +1,64 @@
+setwd(normalizePath(dirname(R.utils::commandArgs(asValues=TRUE)$"f")))
+source('../findNSourceUtils.R')
+
+glm.objectiveFun.test<-
+function(conn) {
+
+    filepath = locate("smalldata/marketing_naRemoved.csv")
+
+    rr=read.csv(filepath)
+    str(rr)
+    dim(rr)
+
+    mfrmr=h2o.uploadFile(conn,filepath,key = "mfrmr")
+    str(mfrmr)
+    myX = 2:13
+    myY = 1 
+    alpha = 1 
+
+    #H2O GLM model  
+    hh=h2o.glm(x=myX,y=myY,data=mfrmr,family="gaussian",nfolds=0, alpha = alpha)
+
+    res_dev = hh@model$deviance
+    obs = nrow(mfrmr)
+    lambda = hh@model$params$best_lambda
+    alpha = hh@model$params$alpha
+    cof = hh@model$normalized_coefficients
+    L1 = sum(abs(cof))
+    L2 = sqrt(sum(cof^2)) 
+    penalty = ( 0.5*(1-alpha)*L2^2 ) + ( alpha*L1 )
+    objective = (res_dev/obs) + ( lambda * penalty )
+
+    # GLMNET Model  
+    gg=glmnet(x=as.matrix(rr[,2:13]),y=(rr[,1]),alpha = alpha,lambda=lambda)
+
+	# Sanity Check whether comparing models built on the same dataset
+	expect_equal( nrow(mfrmr), nrow(rr))
+	expect_equal(gg$nulldev,hh@model$null.deviance)
+
+	res_dev_R = deviance(gg)
+	obs = nrow(mfrmr)
+	cof_R = coef(gg,s= lambda)
+	L1_R = sum(abs(cof_R[,1]))
+	L2_R = sqrt(sum(cof_R[,1]^2))
+	penalty_R = ( 0.5*(1-alpha)*L2_R^2 ) + ( alpha*L1_R )
+	objective_R = (res_dev_R/obs) + ( lambda * penalty_R )
+
+	print(paste("residual deviance from R:  ",res_dev_R, sep = ""))
+	print(paste("residual deviance from H2O:  ",res_dev, sep = ""))
+	print(paste("L1 norm of coefficients from R:  ",L1_R, sep = ""))
+	print(paste("L1 norm of coefficients from H2O:  ",L1, sep = ""))
+	print(paste("L2 norm of coefficients from R:  ",L2_R, sep = ""))
+	print(paste("L2 norm of coefficients from H2O:  ",L2, sep = ""))
+	print(paste("L2 norm of coefficients from R:  ",L2_R, sep = ""))
+	print(paste("penalty on model from R:  ",penalty_R, sep = ""))
+	print(paste("penalty on model from H2O:  ",penalty, sep = ""))
+	print(paste("Objective function for model from R:  ",objective_R, sep = ""))
+	print(paste("Objective function for model from H2O:  ",objective, sep = ""))
+
+	expect_true(objective<=objective_R)
+
+    testEnd()
+}
+doTest("Comapares objective function results from H2O-glm and glmnet: marketing data with no NAs Smalldata", glm.objectiveFun.test)
+