readers

Documentation

Index

• Variables
• type Instance
• type Mnist
  • func (m Mnist) DataFrom(images, labels string) ([]Instance, error)
  • func (m Mnist) PredictDataFrom(filename string) (matrix.Matrix, error)
• type Reader

Variables

var DatasetReaders = map[string]Reader{
	"mnist": Mnist{},
}

DatasetReaders contains a map of the reader types which have strings as keys. The API works as following: the string in the json config file creates an instance of a given reader. Any reader can be created by satisfying the Reader trait and appending the key, value to the map.

Types

type Instance

type Instance struct {
	Image, Label matrix.Matrix
}

Instance is a struct that contains the input and expected output in form of two vectors.

type Mnist

type Mnist struct{}

Mnist reader with two methods that satisfy the Reader interface.

func (Mnist) DataFrom

func (m Mnist) DataFrom(images, labels string) ([]Instance, error)

func (Mnist) PredictDataFrom

func (m Mnist) PredictDataFrom(filename string) (matrix.Matrix, error)

Credits to: https://github.com/sausheong/gonn

type Reader

type Reader interface {
	// DataFrom fetches the data from the two seperate files which contain
	// both the images and labels. This procedure is equivalent when fetching
	// the training and test data.
	DataFrom(images, labels string) ([]Instance, error)
	// DataFromFile is used to perform a predition. It shouldn't have any information
	// about the expected output so it simply parses a vector from an png image.
	PredictDataFrom(filename string) (matrix.Matrix, error)
}

Reader interface is satisfied with Read method which given two filepaths returns a slice of Instances.