Closes #18

Author: AnotherSamWilson

R/bayesOpt.R

@@ -236,20 +236,21 @@ bayesOpt <- function(
 
   # Initialization Setup
   if (missing(initGrid) + missing(initPoints) != 1) stop("Please provide 1 of initGrid or initPoints, but not both.")
-  if (initPoints <= length(bounds)) stop("initPoints must be greater than the number of FUN inputs.")
   if (!missing(initGrid)) {
     setDT(initGrid)
     inBounds <- checkBounds(initGrid,bounds)
     inBounds <- as.logical(apply(inBounds,1,prod))
     if (any(!inBounds)) stop("initGrid not within bounds.")
     optObj$initPars$initialSample <- "User Provided Grid"
+    initPoints <- nrow(initGrid)
   } else {
     initGrid <- randParams(boundsDT, initPoints)
     optObj$initPars$initialSample <- "Latin Hypercube Sampling"
   }
   optObj$initPars$initGrid <- initGrid
   if (nrow(initGrid) <= 2) stop("Cannot initialize with less than 3 samples.")
   optObj$initPars$initPoints <- nrow(initGrid)
+  if (initPoints <= length(bounds)) stop("initPoints must be greater than the number of FUN inputs.")
 
   # Output from FUN is sunk into a temporary file.
   sinkFile <- file()

tests/testthat/test-hyperparameterTuning.R

@@ -75,6 +75,17 @@ testthat::test_that(
       , alpha = c(0,1)
     )
 
+    initGrid <- data.table(
+      max_depth = c(1,1,2,2,3,3,4,4,5)
+      , max_leaves = c(2,3,4,5,6,7,8,9,10)
+      , min_child_weight = seq(bounds$min_child_weight[1],bounds$min_child_weight[2],length.out = 9)
+      , subsample = seq(bounds$subsample[1],bounds$subsample[2],length.out = 9)
+      , colsample_bytree = seq(bounds$colsample_bytree[1],bounds$colsample_bytree[2],length.out = 9)
+      , gamma = seq(bounds$gamma[1],bounds$gamma[2],length.out = 9)
+      , lambda = seq(bounds$lambda[1],bounds$lambda[2],length.out = 9)
+      , alpha = seq(bounds$alpha[1],bounds$alpha[2],length.out = 9)
+    )
+
     optObj <- bayesOpt(
       FUN = scoringFunction
       , bounds = bounds
@@ -86,6 +97,17 @@ testthat::test_that(
 
     expect_equal(nrow(optObj$scoreSummary),13)
 
+    optObj <- bayesOpt(
+      FUN = scoringFunction
+      , bounds = bounds
+      , initGrid = initGrid
+      , iters.n = 4
+      , iters.k = 1
+      , gsPoints = 10
+    )
+
+    expect_equal(nrow(optObj$scoreSummary),13)
+
   }
 
 )