utils

Documentation

Overview

Package utils contains subpackages for independence tests (e.g. Chi-Square) and clustering (e.g. K-Means), and also contains VarData, String Concat and Union-Find.

Index

• Variables
• func AntiLog(l float64) float64
• func Inf(sig int) float64
• func Log(p float64) float64
• func LogProd(p []float64) float64
• func LogSum(p []float64) float64
• func LogSumExp(a []float64) float64
• func LogSumExpPair(l, r float64) float64
• func LogSumLog(v []float64, s []int) (float64, int)
• func LogSumPair(p1, p2 float64) float64
• func Mean(c []int) float64
• func MuSigma(c []int) (float64, float64)
• func PartitionQuantiles(X []int, m int) [][]float64
• func StdDev(c []int) float64
• func StringConcat(s1, s2 string) string
• func Trim(a []float64, c float64) []float64
• func UFVarids(x *UFNode) []int
• type UFNode
  • func Find(x *UFNode) *UFNode
  • func MakeSet(varid int) *UFNode
  • func Union(x, y *UFNode) (*UFNode, int)
• type VarData
  • func NewVarData(varid, categories int, data []int) *VarData

Variables

var (
	// LogZero = ln(0) = -inf
	LogZero float64
	EpsZero float64
)

Functions

func AntiLog

func AntiLog(l float64) float64

AntiLog is the antilog with base e of l. It is equivalent to e raised to the power of l. Thus the following identities apply

antiln(ln(k)) = k
ln(antiln(k)) = k

Returns a float64 that corresponds to the antilog of l.

func Inf

func Inf(sig int) float64

Inf is a typedef for math.Inf.

func Log

func Log(p float64) float64

Log is a typedef for math.Log.

func LogProd

func LogProd(p []float64) float64

LogProd is the log of the product of probabilities given by

prod_i p_i -> sum_i ln(p_i)

Returns a float64 with the resulting log operation. To convert back use utils.AntiLog.

func LogSum

func LogSum(p []float64) float64

LogSum is the log of the sum of probabilities given by

sum_i p_i -> P + ln(sum_i e^(ln(p_i) - P)), where P = max_i ln(p_i)

Returns a float64 with the resulting log operation. To convert back use utils.AntiLog.

func LogSumExp

func LogSumExp(a []float64) float64

LogSumExp takes a slice of floats a={a_1,...,a_n} and computes ln(exp(a_1)+...+exp(a_n)).

func LogSumExpPair

func LogSumExpPair(l, r float64) float64

LogSumExpPair takes two floats l and r and computes ln(l+r). Particular case of LogSumExp.

func LogSumLog

func LogSumLog(v []float64, s []int) (float64, int)

LogSumLog is a function to compute the log of sum of logs.

func LogSumPair

func LogSumPair(p1, p2 float64) float64

LogSumPair is the result of the operation:

ln(p1+p2)=ln(p1)+ln(1+p2/p1)

func Mean

func Mean(c []int) float64

Mean returns the mean of a slice.

func MuSigma

func MuSigma(c []int) (float64, float64)

MuSigma returns both the mean and standard deviation of a slice.

func PartitionQuantiles

func PartitionQuantiles(X []int, m int) [][]float64

PartitionQuantiles takes a slice of values X of a single variable from the dataset and the number m of quantiles to partition the data. Returns a slice of pair of values containing mean and standard deviation (in this order).

func StdDev

func StdDev(c []int) float64

StdDev returns the standard deviation of a slice.

func StringConcat

func StringConcat(s1, s2 string) string

StringConcat concatenates two strings.

func Trim

func Trim(a []float64, c float64) []float64

Trim removes elements from a slice of floats that have the same value as c.

func UFVarids

func UFVarids(x *UFNode) []int

UFVarids returns a slice with all varids in union-find tree x.

Types

type UFNode

type UFNode struct {

	// Pa is Parent of UFNode. Since we use the Path-compression heuristic, this is usually the
	// representative of the set (i.e. the root node).
	Pa *UFNode

	// Children.
	Ch []*UFNode
	// contains filtered or unexported fields
}

UFNode is a Union-Find node on a Union-Find tree. Holds an integer as value. In this case we wish to store the variable ID.

func Find

func Find(x *UFNode) *UFNode

Find returns the representative of x's set.

func MakeSet

func MakeSet(varid int) *UFNode

MakeSet creates a unary set with varid as representative of the resulting set.

func Union

func Union(x, y *UFNode) (*UFNode, int)

Union takes the sets S_1 and S_2, where x is in S_1 and y is in S_2 and unifies S_1 with S_2, returning the union's representative. Returns the union set and 1 if the first argument is the new representative or 2 if the second.

type VarData

type VarData struct {
	// Variable ID.
	Varid int
	// Number of possible instantiations (levels/categories) of Varid.
	Categories int
	// Observed data.
	Data []int
}

VarData is a wrapper struct that contains a variable ID and its observed data. Observed data is data that is to be used in learning. Each data instance i in data (i.e. data[i]) is a variable instantiation.

func NewVarData

func NewVarData(varid, categories int, data []int) *VarData

NewVarData constructs a new VarData. Equivalent to &VarData{varid, categories, data}.