data

Documentation

Index

• func ForEachBatch(datasetSize, batchSize int, callback func(start, end int))
• func GenerateBatches(size, batchFactor int, class func(i int) int) [][]int
• func SplitDataset(size int, splitRatio mat.Float, seed uint64, class func(i int) string) (a []int, b []int)

Functions

func ForEachBatch

func ForEachBatch(datasetSize, batchSize int, callback func(start, end int))

ForEachBatch divides the dataset into batches, returning the start-end of each batch with a callback. This function assumes that the dataset has already been shuffled.

func GenerateBatches

func GenerateBatches(size, batchFactor int, class func(i int) int) [][]int

GenerateBatches generates a list of batches so that the classes distribution among them is approximately the same. The class is given by the callback for each i-th element up to size. The size of each batch depends on number of classes (batchFactor * nClasses). Each batch consists in a list of indices.

func SplitDataset

func SplitDataset(size int, splitRatio mat.Float, seed uint64, class func(i int) string) (a []int, b []int)

SplitDataset splits the dataset into two parts. Each part consists in a list of indices. The split ratio regulates the percentage of the total assigned to `b` so that `a` contains the rest. For example a split ratio of 0.20 means that `b` should contain the 20% of the total and `a` the rest 80%.