base

package
v0.0.0-...-4fa0698 Latest Latest
Warning

This package is not in the latest version of its module.

Go to latest
Published: May 24, 2014 License: MIT Imports: 14 Imported by: 0

Documentation

Overview

Package base provides base interfaces for GoLearn objects to implement. It also provides a raw base for those objects.

Index

Constants

View Source
const (
	// CategoricalType is for Attributes which represent values distinctly.
	CategoricalType = iota
	// Float64Type should be replaced with a FractionalNumeric type [DEPRECATED].
	Float64Type
)

Variables

This section is empty.

Functions

func CheckNewInstancesFromRaw

func CheckNewInstancesFromRaw(attrs []Attribute, rows int, data []float64) error

CheckNewInstancesFromRaw checks whether a call to NewInstancesFromRaw is likely to produce an error-free result.

func ParseCSV

func ParseCSV(filepath string, label int, columns []int) (int, int, []string, []string, []float64)

ParseCSV parses a CSV file and returns the number of columns and rows, the headers, the labels associated with classification, and the data that will be used for training.

func ParseCSVGetRows

func ParseCSVGetRows(filepath string) int

ParseCSVGetRows returns the number of rows in a given file.

func ParseCSVSniffAttributeNames

func ParseCSVSniffAttributeNames(filepath string, hasHeaders bool) []string

ParseCsvSniffAttributeNames returns a slice containing the top row of a given CSV file, or placeholders if hasHeaders is false.

func SaveEstimatorToGob

func SaveEstimatorToGob(path string, e *Estimator)

Serialises an estimator to a provided filepath, in gob format. See http://golang.org/pkg/encoding/gob for further details.

Types

type Attribute

type Attribute interface {
	// Returns the general characterstics of this Attribute .
	// to avoid the overhead of casting
	GetType() int
	// Returns the human-readable name of this Attribute.
	GetName() string
	// Sets the human-readable name of this Attribute.
	SetName(string)
	// Gets a human-readable overview of this Attribute for debugging.
	String() string
	// Converts a value given from a human-readable string into a system
	// representation. For example, a CategoricalAttribute with values
	// ["iris-setosa", "iris-virginica"] would return the float64
	// representation of 0 when given "iris-setosa".
	GetSysValFromString(string) float64
	// Converts a given value from a system representation into a human
	// representation. For example, a CategoricalAttribute with values
	// ["iris-setosa", "iris-viriginica"] might return "iris-setosa"
	// when given 0.0 as the argument.
	GetStringFromSysVal(float64) string
	// Tests for equality with another Attribute. Other Attributes are
	// considered equal if:
	// * They have the same type (i.e. FloatAttribute <> CategoricalAttribute)
	// * They have the same name
	// * If applicable, they have the same categorical values (though not
	//   necessarily in the same order).
	Equals(Attribute) bool
}

Attribute Attributes disambiguate columns of the feature matrix and declare their types.

func ParseCSVGetAttributes

func ParseCSVGetAttributes(filepath string, hasHeaders bool) []Attribute

ParseCSVGetAttributes returns an ordered slice of appropriate-ly typed and named Attributes.

func ParseCSVSniffAttributeTypes

func ParseCSVSniffAttributeTypes(filepath string, hasHeaders bool) []Attribute

ParseCSVSniffAttributeTypes returns a slice of appropriately-typed Attributes.

The type of a given attribute is determined by looking at the first data row of the CSV.

type BaseClassifier

type BaseClassifier struct {
	TrainingData *Instances
}

BaseClassifier stores options common to every classifier.

type BaseEstimator

type BaseEstimator struct {
	Data *mat64.Dense
}

type BaseRegressor

type BaseRegressor struct {
	Data   mat64.Dense
	Name   string
	Labels []float64
}

type CategoricalAttribute

type CategoricalAttribute struct {
	Name string
	// contains filtered or unexported fields
}

CategoricalAttribute is an Attribute implementation which stores discrete string values - useful for representing classes.

func NewCategoricalAttribute

func NewCategoricalAttribute() *CategoricalAttribute

func (*CategoricalAttribute) Equals

func (Attr *CategoricalAttribute) Equals(other Attribute) bool

Equals checks equality against another Attribute.

Two CategoricalAttributes are considered equal if they contain the same values and have the same name. Otherwise, this function returns false.

func (*CategoricalAttribute) GetName

func (Attr *CategoricalAttribute) GetName() string

GetName returns the human-readable name assigned to this attribute.

func (*CategoricalAttribute) GetStringFromSysVal

func (Attr *CategoricalAttribute) GetStringFromSysVal(val float64) string

GetStringFromSysVal returns a human-readable value from the given system-representation value val.

IMPORTANT: This function calls panic() if the value is greater than the length of the array. TODO: Return a user-configurable default instead.

func (*CategoricalAttribute) GetSysVal

func (Attr *CategoricalAttribute) GetSysVal(userVal string) float64

GetSysVal returns the system representation of userVal as an index into the Values slice If the userVal can't be found, it returns -1.

func (*CategoricalAttribute) GetSysValFromString

func (Attr *CategoricalAttribute) GetSysValFromString(rawVal string) float64

GetSysValFromString returns the system representation of rawVal as an index into the Values slice. If rawVal is not inside the Values slice, it is appended.

IMPORTANT: If no system representation yet exists, this functions adds it. If you need to determine whether rawVal exists: use GetSysVal and check for a -1 return value.

Example: if the CategoricalAttribute contains the values ["iris-setosa", "iris-virginica"] and "iris-versicolor" is provided as the argument, the Values slide becomes ["iris-setosa", "iris-virginica", "iris-versicolor"] and 2.00 is returned as the system representation.

func (*CategoricalAttribute) GetType

func (Attr *CategoricalAttribute) GetType() int

GetType returns CategoricalType to avoid casting overhead.

func (*CategoricalAttribute) GetUsrVal

func (Attr *CategoricalAttribute) GetUsrVal(sysVal float64) string

GetUsrVal returns a human-readable representation of the given sysVal.

IMPORTANT: this function doesn't check the boundaries of the array.

func (*CategoricalAttribute) SetName

func (Attr *CategoricalAttribute) SetName(name string)

SetName sets the human-readable name on this attribute.

func (*CategoricalAttribute) String

func (Attr *CategoricalAttribute) String() string

String returns a human-readable summary of this Attribute.

Returns a string containing the list of human-readable values this CategoricalAttribute can take.

type Classifier

type Classifier interface {
	// Takes a set of Instances, copies the class Attribute
	// and constructs a new set of Instances of equivalent
	// length with only the class Attribute and fills it in
	// with predictions.
	Predict(*Instances) *Instances
	// Takes a set of instances and updates the Classifier's
	// internal structures to enable prediction
	Fit(*Instances)
	// Why not make every classifier return a nice-looking string?
	String() string
}

Classifier implementations predict categorical class labels.

type Estimator

type Estimator interface {
	Fit()
}

An object that can ingest some data and train on it.

type FloatAttribute

type FloatAttribute struct {
	Name      string
	Precision int
}

FloatAttribute is an implementation which stores floating point representations of numbers.

func NewFloatAttribute

func NewFloatAttribute() *FloatAttribute

NewFloatAttribute returns a new FloatAttribute with a default precision of 2 decimal places

func (*FloatAttribute) CheckSysValFromString

func (Attr *FloatAttribute) CheckSysValFromString(rawVal string) (float64, error)

CheckSysValFromString confirms whether a given rawVal can be converted into a valid system representation.

func (*FloatAttribute) Equals

func (Attr *FloatAttribute) Equals(other Attribute) bool

Equals tests a FloatAttribute for equality with another Attribute.

Returns false if the other Attribute has a different name or if the other Attribute is not a FloatAttribute.

func (*FloatAttribute) GetName

func (Attr *FloatAttribute) GetName() string

GetName returns this FloatAttribute's human-readable name.

func (*FloatAttribute) GetStringFromSysVal

func (Attr *FloatAttribute) GetStringFromSysVal(rawVal float64) string

GetStringFromSysVal converts a given system value to to a string with two decimal places of precision [TODO: revise this and allow more precision].

func (*FloatAttribute) GetSysVal

func (Attr *FloatAttribute) GetSysVal(userVal float64) float64

GetSysVal returns the system representation of userVal.

Because FloatAttribute represents float64 types, this just returns its argument.

func (*FloatAttribute) GetSysValFromString

func (Attr *FloatAttribute) GetSysValFromString(rawVal string) float64

GetSysValFromString parses the given rawVal string to a float64 and returns it.

float64 happens to be a 1-to-1 mapping to the system representation. IMPORTANT: This function panic()s if rawVal is not a valid float. Use CheckSysValFromString to confirm.

func (*FloatAttribute) GetType

func (Attr *FloatAttribute) GetType() int

GetType returns Float64Type.

func (*FloatAttribute) GetUsrVal

func (Attr *FloatAttribute) GetUsrVal(sysVal float64) float64

GetUsrVal returns the user representation of sysVal.

Because FloatAttribute represents float64 types, this just returns its argument.

func (*FloatAttribute) SetName

func (Attr *FloatAttribute) SetName(name string)

SetName sets this FloatAttribute's human-readable name.

func (*FloatAttribute) String

func (Attr *FloatAttribute) String() string

String returns a human-readable summary of this Attribute. e.g. "FloatAttribute(Sepal Width)"

type Instances

type Instances struct {
	Rows       int
	Cols       int
	ClassIndex int
	// contains filtered or unexported fields
}

Instances represents a grid of numbers (typed by Attributes) stored internally in mat.DenseMatrix as float64's. See docs/instances.md for more information.

func InstancesTrainTestSplit

func InstancesTrainTestSplit(src *Instances, prop float64) [2](*Instances)

InstancesTrainTestSplit takes a given Instances (src) and a train-test fraction (prop) and returns an array of two new Instances, one containing approximately that fraction and the other containing what's left.

IMPORTANT: this function is only meaningful when prop is between 0.0 and 1.0. Using any other values may result in odd behaviour.

func NewInstances

func NewInstances(attrs []Attribute, rows int) *Instances

NewInstances returns a preallocated Instances structure with some helful values pre-filled.

func NewInstancesFromDense

func NewInstancesFromDense(attrs []Attribute, rows int, mat *mat64.Dense) *Instances

NewInstancesFromDense creates a set of Instances from a mat64.Dense matrix

func NewInstancesFromRaw

func NewInstancesFromRaw(attrs []Attribute, rows int, data []float64) *Instances

NewInstancesFromRaw wraps a slice of float64 numbers in a mat64.Dense structure, reshaping it with the given number of rows and representing it with the given attrs (Attribute slice)

IMPORTANT: if the |attrs| * |rows| value doesn't equal len(data) then panic()s may occur. Use CheckNewInstancesFromRaw to confirm.

func ParseCSVToInstances

func ParseCSVToInstances(filepath string, hasHeaders bool) (instances *Instances, err error)

ParseCSVToInstances reads the CSV file given by filepath and returns the read Instances.

func (*Instances) CountAttrValues

func (inst *Instances) CountAttrValues(a Attribute) map[string]int

CountAttrValues returns the distribution of values of a given Attribute. IMPORTANT: calls panic() if the attribute index of a cannot be determined. Call GetAttrIndex(a) and check for a -1 return value.

func (*Instances) CountClassValues

func (inst *Instances) CountClassValues() map[string]int

CountClassValues returns the class distribution of this Instances set

func (*Instances) DecomposeOnAttributeValues

func (inst *Instances) DecomposeOnAttributeValues(at Attribute) map[string]*Instances

DecomposeOnAttributeValues divides the instance set depending on the value of a given Attribute, constructs child instances, and returns them in a map keyed on the string value of that Attribute. IMPORTANT: calls panic() if the attribute index of at cannot be determined. Use GetAttrIndex(at) and check for a non-zero return value.

func (*Instances) Equal

func (inst *Instances) Equal(other *Instances) bool

Equal checks whether a given Instance set is exactly the same as another: same size and same values (as determined by the Attributes)

IMPORTANT: does not explicitly check if the Attributes are considered equal.

func (*Instances) GeneratePredictionVector

func (inst *Instances) GeneratePredictionVector() *Instances

GeneratePredictionVector generates a new set of Instances with the same number of rows, but only this Instance set's class Attribute.

func (*Instances) Get

func (inst *Instances) Get(row int, col int) float64

Get returns the system representation (float64) of the value stored at the given row and col coordinate.

func (*Instances) GetAttr

func (inst *Instances) GetAttr(attrIndex int) Attribute

GetAttr returns information about an attribute at given index in the attributes slice.

func (*Instances) GetAttrIndex

func (inst *Instances) GetAttrIndex(of Attribute) int

GetAttrIndex returns the offset of a given Attribute `a' to an index in the attributes slice

func (*Instances) GetAttrStr

func (inst *Instances) GetAttrStr(row int, attr int) string

GetAttrStr returns a human-readable string value stored in column `attr' and row `row', as determined by the appropriate Attribute function.

func (*Instances) GetAttributeCount

func (inst *Instances) GetAttributeCount() int

GetAttributeCount returns the number of attributes represented.

func (*Instances) GetClass

func (inst *Instances) GetClass(row int) string

GetClass returns the string representation of the given row's class, as determined by the Attribute at the ClassIndex position from GetAttr

func (*Instances) GetClassAttr

func (Inst *Instances) GetClassAttr() Attribute

func (*Instances) GetClassAttrPtr

func (Inst *Instances) GetClassAttrPtr() *Attribute

func (*Instances) GetClassDistribution

func (inst *Instances) GetClassDistribution() map[string]int

GetClassDist returns a map containing the count of each class type (indexed by the class' string representation)

func (*Instances) GetClassDistributionAfterSplit

func (inst *Instances) GetClassDistributionAfterSplit(at Attribute) map[string]map[string]int

func (*Instances) GetRowVector

func (inst *Instances) GetRowVector(row int) []float64

GetRowVector returns a row of system representation values at the given row index.

func (*Instances) GetRowVectorWithoutClass

func (inst *Instances) GetRowVectorWithoutClass(row int) []float64

GetRowVector returns a row of system representation values at the given row index, excluding the class attribute

func (*Instances) ReplaceAttr

func (inst *Instances) ReplaceAttr(index int, a Attribute)

ReplaceAttr overwrites the attribute at `index' with `a'

func (*Instances) RowStr

func (inst *Instances) RowStr(row int) string

RowStr returns a human-readable representation of a given row.

func (*Instances) SampleWithReplacement

func (inst *Instances) SampleWithReplacement(size int) *Instances

SampleWithReplacement returns a new set of Instances of size `size' containing random rows from this set of Instances.

IMPORTANT: There's a high chance of seeing duplicate rows whenever size is close to the row count.

func (*Instances) SelectAttributes

func (inst *Instances) SelectAttributes(attrs []Attribute) *Instances

SelectAttributes returns a new instance set containing the values from this one with only the Attributes specified

func (*Instances) Set

func (inst *Instances) Set(row int, col int, val float64)

Set sets the system representation (float64) to val at the given row and column coordinate.

func (*Instances) SetAttrStr

func (inst *Instances) SetAttrStr(row int, attr int, val string)

SetAttrStr sets the system-representation value of row in column attr to value val, implicitly converting the string to system-representation via the appropriate Attribute function.

func (*Instances) Shuffle

func (inst *Instances) Shuffle()

Shuffle randomizes the row order in place

func (*Instances) Sort

func (inst *Instances) Sort(direction SortDirection, attrs []int)

Sort does an in-place radix sort of Instances, using SortDirection direction (Ascending or Descending) with attrs as a slice of Attribute indices that you want to sort by.

IMPORTANT: Radix sort is not stable, so ordering outside the attributes used for sorting is arbitrary.

func (*Instances) String

func (inst *Instances) String() string

type Model

type Model interface {
	Score()
}

An supervised learning object, that is possible of scoring accuracy against a test set.

type Predictor

type Predictor interface {
	Predict()
}

An object that provides predictions.

type SortDirection

type SortDirection int

SortDirection specifies sorting direction...

const (
	// Descending says that Instances should be sorted high to low...
	Descending SortDirection = 1
	// Ascending states that Instances should be sorted low to high...
	Ascending SortDirection = 2
)

Jump to

Keyboard shortcuts

? : This menu
/ : Search site
f or F : Jump to
y or Y : Canonical URL