generic.cv.R

[whizzalan]

When I used generic.cv for madlib.elnet, I got some problem, as follow.

x <- matrix(rnorm(100*20),100,20)
y <- rbinom(100,1,0.5)
dat <- data.frame(x, y)
delete("eldata")
z <- as.db.data.frame(dat, "eldata", verbose = FALSE)

g <- generic.cv(
  train = function (data, alpha, lambda) {
    madlib.elnet(y ~ ., data = data, family = "binomial",
                 alpha = alpha, lambda = lambda)
  },
  predict = predict, #function(fit,newdata) {predict(fit,newdata,type="response")} ,
  metric = function (predicted, data) {
    lk(mean((data$y - predicted)^2))
  },
  data = z,
  params = list(alpha=1, lambda=seq(0,0.2,0.1)),
  k = 3, find.min = TRUE)

Computation in-database ... Cutting the data row-wise into 3 pieces ... Running on fold 1 now ... parameters 1, 0 ... parameters 1, 0.1 ... parameters 1, 0.2 ... Running on fold 2 now ... parameters 1, 0 ... parameters 1, 0.1 ... parameters 1, 0.2 ... Running on fold 3 now ... parameters 1, 0 ... parameters 1, 0.1 ... parameters 1, 0.2 ... Done. Fitting the best model using the whole data set ... Executing in database connection 1:

select madlib.elastic_net_train('"eldata"', 'madlib_temp_5d38a1b6_056d_a5aba2_90e7cb4f218a', '("y")::boolean', 'array["X1","X2","X3","X4","X5","X6","X7","X8","X9","X10","X11","X12","X13","X14","X15","X16","X17","X18","X19","X20"]', 'binomial', , , TRUE, NULL, 'fista', 'use_active_set = f', NULL, 100, 1e-04)

[walkingsparrow]

This seems to be a simple bug, where three "," are seen in the problematic query.

[whizzalan]

How do I modify the source code?

[fmcquillan99]

MADlib dev team will have a look at this and update this thread. Thanks, Frank

[whizzalan]

I think the problem is the function of generic.cv in R not MADlib. Is there any resource of how to modify the source code with R ?

[sziegler11-zz]

In the definition of the metric function, "predicted" should be cast to from boolean to integers via the "as.integer" function. Without this, the database is trying to subtract booleans with numeric values and returning NA.