base

package
v0.1.1 Latest Latest
Warning

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

 Go to latest Published: Nov 24, 2019 License: Apache-2.0 Imports: 9 Imported by: 2

Documentation

Overview

Package base provides base data structures and utils.

The base data structures and utils include:

  • Parallel Scheduler
  • Hyper-parameters Management
  • Random Generator
  • Similarity Metrics
  • Sparse Data Structures

Index

Constants

View Source 
const (
	Basic    string = "basic"    // Basic KNN
	Centered string = "centered" // KNN with centered ratings
	ZScore   string = "z_score"  // KNN with standardized ratings
	Baseline string = "baseline" // KNN with baseline ratings
)

Predefined values for hyper-parameter Type.

View Source 
const (
	SGDOptimizer string = "sgd" // Fit model (FM) with stochastic gradient descent.
	BPROptimizer string = "bpr" // Fit model (FM) with bayesian personal ranking.
)

Predefined values for hyper-parameter Optimizer.

View Source 
const (
	Pearson string = "pearson" // Pearson similarity
	Cosine  string = "cosine"  // Cosine similarity
	MSD     string = "msd"     // MSD similarity
)

Predefined values for hyper-parameter Similarity.

View Source 
const NotId = -1

NotId represents an ID doesn't exist.

Variables

This section is empty.

Functions

func Concatenate 

func Concatenate(vectors ...[]int) []int

Concatenate merges vectors of integers to one.

func CosineSimilarity 

func CosineSimilarity(a, b *MarginalSubSet) float64

CosineSimilarity computes the cosine similarity between a pair of vectors.

func FillZeroMatrix 

func FillZeroMatrix(mat [][]float64)

FillZeroMatrix fills a matrix with zeros.

func FillZeroVector 

func FillZeroVector(vec []float64)

FillZeroVector fills a vector with zeros.

func ImplicitSimilarity 

func ImplicitSimilarity(a, b *MarginalSubSet) float64

ImplicitSimilarity computes similarity between two vectors with implicit feedback.

func MSDSimilarity 

func MSDSimilarity(a, b *MarginalSubSet) float64

MSDSimilarity computes the Mean Squared Difference similarity between a pair of vectors.

func Max 

func Max(a []int) int

Max finds the maximum in a vector of integers. Panic if the slice is empty.

func Min 

func Min(a []int) int

Min finds the minimum in a vector of integers. Panic if the slice is empty.

func NewMatrix 

func NewMatrix(row, col int) [][]float64

NewMatrix creates a matrix.

func NewMatrixInt 

func NewMatrixInt(row, col int) [][]int

NewMatrixInt creates a matrix of integers.

func Parallel 

func Parallel(nTask int, nJob int, worker func(begin, end int))

Parallel schedules and runs tasks in parallel. nTask is the number of tasks. nJob is the number of executors. worker is the executed function which passed a range of task IDs (begin, end).

func ParallelFor 

func ParallelFor(begin, end int, worker func(i int))

ParallelFor runs for loop in parallel.

func ParallelForSum 

func ParallelForSum(begin, end int, worker func(i int) float64) float64

ParallelForSum runs for loop in parallel.

func ParallelMean 

func ParallelMean(nTask int, nJob int, worker func(begin, end int) (sum float64)) float64

ParallelMean schedules and runs tasks in parallel, then returns the mean of returned values. nJob is the number of executors. worker is the executed function which passed a range of task IDs (begin, end) and returns a double value.

func PearsonSimilarity 

func PearsonSimilarity(a, b *MarginalSubSet) float64

PearsonSimilarity computes the absolute Pearson correlation coefficient between a pair of vectors.

Types

type FuncSimilarity 

type FuncSimilarity func(a, b *MarginalSubSet) float64

FuncSimilarity computes the similarity between a pair of vectors.

type Indexer 

type Indexer struct {
	Indices map[int]int // sparse ID -> dense index
	IDs     []int       // dense index -> sparse ID
}

Indexer manages the map between sparse IDs and dense indices. A sparse ID is a user ID or item ID. The dense index is the internal user index or item index optimized for faster parameter access and less memory usage.

func NewIndexer 

func NewIndexer() *Indexer

NewIndexer creates a Indexer.

func (*Indexer) Add 

func (set *Indexer) Add(ID int)

Add adds a new ID to the indexer.

func (*Indexer) Len 

func (set *Indexer) Len() int

Len returns the number of indexed IDs.

func (*Indexer) ToID 

func (set *Indexer) ToID(index int) int

ToID converts a dense index to a sparse ID.

func (*Indexer) ToIndex 

func (set *Indexer) ToIndex(ID int) int

ToIndex converts a sparse ID to a dense index.

type MarginalSubSet 

type MarginalSubSet struct {
	Indexer *Indexer  // the indexer
	Indices []int     // the full list of indices
	Values  []float64 // the full list of values
	SubSet  []int     // indices of the subset
}

MarginalSubSet constructs a subset over a list of IDs, indices and values.

func NewMarginalSubSet 

func NewMarginalSubSet(indexer *Indexer, indices []int, values []float64, subset []int) *MarginalSubSet

NewMarginalSubSet creates a MarginalSubSet.

func (*MarginalSubSet) Contain 

func (set *MarginalSubSet) Contain(id int) bool

Contain returns true am ID existed in the subset.

func (*MarginalSubSet) Count 

func (set *MarginalSubSet) Count() int

Count gets the size of marginal subset.

func (*MarginalSubSet) ForEach 

func (set *MarginalSubSet) ForEach(f func(i, id int, value float64))

ForEach iterates items in the subset with IDs.

func (*MarginalSubSet) ForEachIndex 

func (set *MarginalSubSet) ForEachIndex(f func(i, index int, value float64))

ForEachIndex iterates items in the subset with indices.

func (*MarginalSubSet) ForIntersection 

func (set *MarginalSubSet) ForIntersection(other *MarginalSubSet, f func(id int, a, b float64))

ForIntersection iterates items in the intersection of two subsets. The method find items with common indices in linear time.

func (*MarginalSubSet) GetID 

func (set *MarginalSubSet) GetID(i int) int

GetID returns the ID of i-th item.

func (*MarginalSubSet) GetIndex 

func (set *MarginalSubSet) GetIndex(i int) int

GetIndex returns the index of i-th item.

func (*MarginalSubSet) Len 

func (set *MarginalSubSet) Len() int

Len returns the number of items.

func (*MarginalSubSet) Less 

func (set *MarginalSubSet) Less(i, j int) bool

Less compares two items.

func (*MarginalSubSet) Mean 

func (set *MarginalSubSet) Mean() float64

Mean of ratings in the subset.

func (*MarginalSubSet) Swap 

func (set *MarginalSubSet) Swap(i, j int)

Swap two items.

type MaxHeap 

type MaxHeap struct {
	Elem  []interface{} // store elements
	Score []float64     // store scores
	K     int           // the size of heap
}

MaxHeap is designed for store K maximal elements. Heap is used to reduce time complexity and memory complexity in top-K searching.

func NewMaxHeap 

func NewMaxHeap(k int) *MaxHeap

NewMaxHeap creates a MaxHeap.

func (*MaxHeap) Add 

func (maxHeap *MaxHeap) Add(elem interface{}, score float64)

Add a new element to the MaxHeap.

func (*MaxHeap) Len 

func (maxHeap *MaxHeap) Len() int

Len returns the size of heap. It is a method of heap.Interface.

func (*MaxHeap) Less 

func (maxHeap *MaxHeap) Less(i, j int) bool

Less returns true if the score of i-th item is less than the score of j-th item. It is a method of heap.Interface.

func (*MaxHeap) Pop 

func (maxHeap *MaxHeap) Pop() interface{}

Pop the last item (the element with minimal score) in the MaxHeap. It is a method of heap.Interface.

func (*MaxHeap) Push 

func (maxHeap *MaxHeap) Push(x interface{})

Push a neighbors into the MaxHeap. It is a method of heap.Interface.

func (*MaxHeap) Swap 

func (maxHeap *MaxHeap) Swap(i, j int)

Swap the i-th item with the j-th item. It is a method of heap.Interface.

func (*MaxHeap) ToSorted 

func (maxHeap *MaxHeap) ToSorted() ([]interface{}, []float64)

ToSorted returns a sorted slice od elements in the heap.

type ParamName 

type ParamName string

ParamName is the type of hyper-parameter names. 

const (
	Lr            ParamName = "Lr"            // learning rate
	Reg           ParamName = "Reg"           // regularization strength
	NEpochs       ParamName = "NEpochs"       // number of epochs
	NFactors      ParamName = "NFactors"      // number of factors
	RandomState   ParamName = "RandomState"   // random state (seed)
	UseBias       ParamName = "UseBias"       // use bias
	InitMean      ParamName = "InitMean"      // mean of gaussian initial parameter
	InitStdDev    ParamName = "InitStdDev"    // standard deviation of gaussian initial parameter
	InitLow       ParamName = "InitLow"       // lower bound of uniform initial parameter
	InitHigh      ParamName = "InitHigh"      // upper bound of uniform initial parameter
	NUserClusters ParamName = "NUserClusters" // number of user cluster
	NItemClusters ParamName = "NItemClusters" // number of item cluster
	Type          ParamName = "Type"          // type for KNN
	UserBased     ParamName = "UserBased"     // user based if true. otherwise item based.
	Similarity    ParamName = "Similarity"    // similarity metrics
	K             ParamName = "K"             // number of neighbors
	MinK          ParamName = "MinK"          // least number of neighbors
	Optimizer     ParamName = "Optimizer"     // optimizer for optimization (SGD/ALS/BPR)
	Shrinkage     ParamName = "Shrinkage"     // shrinkage strength of similarity
	Alpha         ParamName = "Alpha"         // alpha value, depend on context
)

Predefined hyper-parameter names

type Params 

type Params map[ParamName]interface{}

Params stores hyper-parameters for an model. It is a map between strings (names) and interface{}s (values). For example, hyper-parameters for SVD is given by: 

 base.Params{
		base.Lr:       0.007,
		base.NEpochs:  100,
		base.NFactors: 80,
		base.Reg:      0.1,
	}

func (Params) Copy 

func (parameters Params) Copy() Params

Copy hyper-parameters.

func (Params) GetBool 

func (parameters Params) GetBool(name ParamName, _default bool) bool

GetBool gets a bool parameter by name. Returns _default if not exists or type doesn't match.

func (Params) GetFloat64 

func (parameters Params) GetFloat64(name ParamName, _default float64) float64

GetFloat64 gets a float parameter by name. Returns _default if not exists or type doesn't match. The type will be converted if given int.

func (Params) GetInt 

func (parameters Params) GetInt(name ParamName, _default int) int

GetInt gets a integer parameter by name. Returns _default if not exists or type doesn't match.

func (Params) GetInt64 

func (parameters Params) GetInt64(name ParamName, _default int64) int64

GetInt64 gets a int64 parameter by name. Returns _default if not exists or type doesn't match. The type will be converted if given int.

func (Params) GetString 

func (parameters Params) GetString(name ParamName, _default string) string

GetString gets a string parameter. Returns _default if not exists or type doesn't match.

func (Params) Merge 

func (parameters Params) Merge(params Params) Params

Merge another group of hyper-parameters to current group of hyper-parameters.

type RandomGenerator 

type RandomGenerator struct {
	*rand.Rand
}

RandomGenerator is the random generator for gorse.

func NewRandomGenerator 

func NewRandomGenerator(seed int64) RandomGenerator

NewRandomGenerator creates a RandomGenerator.

func (RandomGenerator) NewNormalMatrix 

func (rng RandomGenerator) NewNormalMatrix(row, col int, mean, stdDev float64) [][]float64

NewNormalMatrix makes a matrix filled with normal random floats.

func (RandomGenerator) NewNormalVector 

func (rng RandomGenerator) NewNormalVector(size int, mean, stdDev float64) []float64

NewNormalVector makes a vec filled with normal random floats.

func (RandomGenerator) NewUniformMatrix 

func (rng RandomGenerator) NewUniformMatrix(row, col int, low, high float64) [][]float64

NewUniformMatrix makes a matrix filled with uniform random floats.

func (RandomGenerator) NewUniformVector 

func (rng RandomGenerator) NewUniformVector(size int, low, high float64) []float64

NewUniformVector makes a vec filled with uniform random floats,

func (RandomGenerator) NewUniformVectorInt 

func (rng RandomGenerator) NewUniformVectorInt(size, low, high int) []int

NewUniformVectorInt makes a vec filled with uniform random integers.

type RuntimeOptions 

type RuntimeOptions struct {
	Verbose bool // Verbose switch
	FitJobs int  // Number of jobs for model fitting
	CVJobs  int  // Number of jobs for cross validation
}

RuntimeOptions defines options used for runtime.

func (*RuntimeOptions) GetCVJobs 

func (options *RuntimeOptions) GetCVJobs() int

GetCVJobs returns the number of concurrent jobs for cross validation.

func (*RuntimeOptions) GetFitJobs 

func (options *RuntimeOptions) GetFitJobs() int

GetFitJobs returns the number of concurrent jobs for model fitting.

func (*RuntimeOptions) GetVerbose 

func (options *RuntimeOptions) GetVerbose() bool

GetVerbose returns the indicator of verbose.

func (*RuntimeOptions) Log 

func (options *RuntimeOptions) Log(v ...interface{})

Log to logs.

func (*RuntimeOptions) Logf 

func (options *RuntimeOptions) Logf(format string, v ...interface{})

Logf to logs with format.

func (*RuntimeOptions) Logln 

func (options *RuntimeOptions) Logln(v ...interface{})

Logln to logs with newline.

type SparseVector 

type SparseVector struct {
	Indices []int
	Values  []float64
	Sorted  bool
}

SparseVector is the data structure for the sparse vector.

func NewDenseSparseMatrix 

func NewDenseSparseMatrix(row int) []*SparseVector

NewDenseSparseMatrix creates an array of SparseVectors.

func NewSparseVector 

func NewSparseVector() *SparseVector

NewSparseVector creates a SparseVector.

func (*SparseVector) Add 

func (vec *SparseVector) Add(index int, value float64)

Add a new item.

func (*SparseVector) ForEach 

func (vec *SparseVector) ForEach(f func(i, index int, value float64))

ForEach iterates items in the sparse vector.

func (*SparseVector) ForIntersection 

func (vec *SparseVector) ForIntersection(other *SparseVector, f func(index int, a, b float64))

ForIntersection iterates items in the intersection of two vectors. The method sorts two vectors by indices first, then find common indices in linear time.

func (*SparseVector) Len 

func (vec *SparseVector) Len() int

Len returns the number of items.

func (*SparseVector) Less 

func (vec *SparseVector) Less(i, j int) bool

Less returns true if the index of i-th item is less than the index of j-th item.

func (*SparseVector) SortIndex 

func (vec *SparseVector) SortIndex()

SortIndex sorts items by indices.

func (*SparseVector) Swap 

func (vec *SparseVector) Swap(i, j int)

Swap two items.

type StringIndexer 

type StringIndexer struct {
	Indices map[string]int
	Names   []string
}

StringIndexer manages the map between names and indices. The index is the internal index optimized for faster parameter access and less memory usage.

func NewStringIndexer 

func NewStringIndexer() *StringIndexer

NewStringIndexer creates a StringIndexer.

func (*StringIndexer) Add 

func (set *StringIndexer) Add(name string)

Add adds a new ID to the indexer.

func (*StringIndexer) Len 

func (set *StringIndexer) Len() int

Len returns the number of indexed IDs.

func (*StringIndexer) ToIndex 

func (set *StringIndexer) ToIndex(name string) int

ToIndex converts a sparse ID to a dense index.

func (*StringIndexer) ToName 

func (set *StringIndexer) ToName(index int) string

ToName converts an index to a name.

Source Files

View all Source files

Jump to

Keyboard shortcuts

? : This menu
/ : Search site
f or F : Jump to
y or Y : Canonical URL
go.dev uses cookies from Google to deliver and enhance the quality of its services and to analyze traffic. Learn more.