BeforeEach(func() {
trainingRatio = 0.80
})
Context("when the dataset has no rows", func() {
It("errors", func() {
Ω(err).Should(HaveOccurred())
})
})
Context("when the dataset has rows and the ratio is valid", func() {
var expectedTrainingTargets, expectedTestTargets []slice.Slice
BeforeEach(func() {
for i := 0; i < 20; i++ {
originalSet.AddRowFromStrings([]string{strconv.Itoa(i)})
}
expectedTrainingTargets = makeSingleFloatTargets(16, 13, 1, 2, 19, 6, 5, 14, 10, 11, 4, 15, 8, 3, 9)
expectedTestTargets = makeSingleFloatTargets(12, 7, 17, 0, 18)
})
It("splits the dataset into two according to the given ratio", func() {
Ω(err).ShouldNot(HaveOccurred())
Ω(trainingSet.NumRows()).Should(Equal(len(expectedTrainingTargets)))
for i := 0; i < trainingSet.NumRows(); i++ {
trainingRow, rowErr := trainingSet.Row(i)
Ω(rowErr).ShouldNot(HaveOccurred())
Ω(trainingRow.Target().Equals(expectedTrainingTargets[i])).Should(BeTrue())
}
_, err = ds.Row(-1)
Ω(err).Should(HaveOccurred())
_, err = ds.Row(1)
Ω(err).Should(HaveOccurred())
})
})
Context("When adding a row", func() {
var newRow row.Row
var err error
Context("When the row's length is incorrect", func() {
BeforeEach(func() {
err = ds.AddRowFromStrings([]string{"0.0", "hi", "mom", "94"})
})
It("Returns an error", func() {
Ω(err).Should(HaveOccurred())
})
It("Has 0 rows", func() {
Ω(ds.NumRows()).To(BeZero())
})
Context("When getting a row", func() {
It("Returns an error", func() {
_, err = ds.Row(0)
Ω(err).Should(HaveOccurred())
0.3,
0.3,
100,
lossGradient,
)
Ω(err).ShouldNot(HaveOccurred())
})
Context("Given a dataset with non-float features", func() {
BeforeEach(func() {
columnTypes, err := columntype.StringsToColumnTypes([]string{"x", "1.0"})
Ω(err).ShouldNot(HaveOccurred())
trainingSet = dataset.NewDataset([]int{0}, []int{1}, columnTypes)
err = trainingSet.AddRowFromStrings([]string{"hi", "24"})
Ω(err).ShouldNot(HaveOccurred())
})
It("Returns an error", func() {
err := estimator.Train(trainingSet)
Ω(err).Should(BeAssignableToTypeOf(gdeErrors.NonFloatFeaturesError{}))
})
})
Context("Given a dataset with a non-float target", func() {
BeforeEach(func() {
columnTypes, err := columntype.StringsToColumnTypes([]string{"x", "1.0"})
Ω(err).ShouldNot(HaveOccurred())
trainingSet = dataset.NewDataset([]int{1}, []int{0}, columnTypes)
It("Returns an error", func() {
err := kNNClassifier.Train(trainingData)
Ω(err).Should(HaveOccurred())
Ω(err).Should(BeAssignableToTypeOf(knnerrors.NonFloatFeaturesTrainingSetError{}))
})
})
Context("When the dataset is valid", func() {
BeforeEach(func() {
columnTypes, err := columntype.StringsToColumnTypes([]string{"hi", "0", "0"})
Ω(err).ShouldNot(HaveOccurred())
trainingData = dataset.NewDataset([]int{1, 2}, []int{0}, columnTypes)
err = trainingData.AddRowFromStrings([]string{"hi", "0", "0"})
Ω(err).ShouldNot(HaveOccurred())
})
It("Doesn't return an error", func() {
err := kNNClassifier.Train(trainingData)
Ω(err).ShouldNot(HaveOccurred())
})
})
})
Describe("Classify", func() {
var testRow row.Row
var emptyTarget slice.Slice
var columnTypes []columntype.ColumnType
var err error
It("Returns an error", func() {
err := linearRegressor.Train(trainingData)
Ω(err).Should(HaveOccurred())
Ω(err).Should(BeAssignableToTypeOf(linearerrors.NoFeaturesError{}))
})
})
Context("When the dataset is valid", func() {
BeforeEach(func() {
columnTypes, err := columntype.StringsToColumnTypes([]string{"1.2", "3.4", "5.6"})
Ω(err).ShouldNot(HaveOccurred())
trainingData = dataset.NewDataset([]int{1, 2}, []int{0}, columnTypes)
err = trainingData.AddRowFromStrings([]string{"1.2", "3.0", "0.5"})
Ω(err).ShouldNot(HaveOccurred())
})
It("Doesn't return an error", func() {
err := linearRegressor.Train(trainingData)
Ω(err).ShouldNot(HaveOccurred())
})
})
})
Describe("Predict", func() {
var testRow row.Row
var emptyTarget slice.Slice
var columnTypes []columntype.ColumnType
var err error