Packages

package dataset

Ordering
  1. Alphabetic
Visibility
  1. Public
  2. All

Type Members

  1. class LightGBMDataset extends AutoCloseable

    Represents a LightGBM dataset.

    Represents a LightGBM dataset. Wraps the native implementation.

  2. class PeekingIterator[T] extends Iterator[T]
  3. case class SampledData(numRows: Int, numCols: Int) extends Product with Serializable

    SampledData: Encapsulates the sampled data need to initialize a LightGBM dataset.

    SampledData: Encapsulates the sampled data need to initialize a LightGBM dataset. . LightGBM expects sampled data to be an array of vectors, where each feature column has a sparse representation of non-zero values (i.e. indexes and data vector). It also needs a #features sized array of element count per feature to know how long each column is. . Since we create sampled data as a self-contained set with ONLY sampled data and nothing else, the indexes are trivial (0 until #elements). We don't need to maintain original raw indexes. LightGBM only uses this data to get distributions, and does not care about raw row indexes. . This class manages keeping all the indexing in sync so callers can just push rows of data into it and retrieve the resulting pointers at the end. . Note: sample data row count is not expected to exceed max(Int), so we index with Ints.

Value Members

  1. object DatasetUtils
  2. object ReferenceDatasetUtils

Ungrouped