Documentation
¶
Index ¶
- func ConvertToMap(wv WordVectors, vocab map[string]int) map[string][]float64
- func ReadTrainingData(filePath string) ([]string, error)
- type SimpleWord2Vec
- func (sw2v *SimpleWord2Vec) Backpropagate(output, context []float64)
- func (sw2v *SimpleWord2Vec) CalculateLoss(output, context []float64) float64
- func (sw2v *SimpleWord2Vec) ForwardPass(words []string) []float64
- func (sw2v *SimpleWord2Vec) InitializeWeights()
- func (sw2v *SimpleWord2Vec) SaveModel(filename string) error
- func (sw2v *SimpleWord2Vec) Train(trainingData []string)
- func (sw2v *SimpleWord2Vec) TrainSentenceContext(sentences []string) map[string][]float64
- type TrainingData
- type Vector
- type WordVectors
Constants ¶
This section is empty.
Variables ¶
This section is empty.
Functions ¶
func ConvertToMap ¶
func ConvertToMap(wv WordVectors, vocab map[string]int) map[string][]float64
convertToMap converts WordVectors to a map[string][]float64.
func ReadTrainingData ¶
ReadTrainingData reads the training data from a JSON file.
Types ¶
type SimpleWord2Vec ¶
type SimpleWord2Vec struct {
// Core word representation
Vocabulary map[string]int
WordVectors map[int][]float64
VectorSize int
NgramVectors map[string][]float64
VocabSize int
// Context and semantic information
ContextEmbeddings map[string][]float64
ContextLabels map[string]string
SentenceTags map[string][]string
UNKToken string
// Hyperparameters and network configuration
LearningRate float64
MaxGrad float64
HiddenSize int
SimilarityThreshold float64
Window int
Epochs int
NegativeSamples int
UseCBOW bool
NgramSize int
// Weights and biases (neural network parameters)
Weights [][]float64
Biases [][]float64
Ann *g.ANN
//MinWordFrequency
MinWordFrequency int
}
func LoadModel ¶
func LoadModel(filename string) (*SimpleWord2Vec, error)
LoadModel loads a trained model from a GOB file.
func TrainWord2VecModel ¶
func TrainWord2VecModel(trainingDataPath, modelSavePath string, vectorSize, epochs, window, negativeSamples, minWordFrequency int, useCBOW bool) (*SimpleWord2Vec, error)
TrainWord2VecModel initializes, trains, and saves a SimpleWord2Vec model.
func (*SimpleWord2Vec) Backpropagate ¶
func (sw2v *SimpleWord2Vec) Backpropagate(output, context []float64)
backpropagate performs backpropagation and updates weights and biases.
func (*SimpleWord2Vec) CalculateLoss ¶
func (sw2v *SimpleWord2Vec) CalculateLoss(output, context []float64) float64
calculateLoss calculates the Mean Squared Error (MSE) loss.
func (*SimpleWord2Vec) ForwardPass ¶
func (sw2v *SimpleWord2Vec) ForwardPass(words []string) []float64
forwardPass performs the forward pass of the RNN.
func (*SimpleWord2Vec) InitializeWeights ¶
func (sw2v *SimpleWord2Vec) InitializeWeights()
InitializeWeights initializes the weights and biases of the RNN.
func (*SimpleWord2Vec) SaveModel ¶
func (sw2v *SimpleWord2Vec) SaveModel(filename string) error
SaveModel saves the trained model to a GOB file.
func (*SimpleWord2Vec) Train ¶
func (sw2v *SimpleWord2Vec) Train(trainingData []string)
Train implements a simplified Skip-gram model for training word embeddings.
func (*SimpleWord2Vec) TrainSentenceContext ¶
func (sw2v *SimpleWord2Vec) TrainSentenceContext(sentences []string) map[string][]float64
type TrainingData ¶
type TrainingData struct {
Sentences []struct {
Tokens []string `json:"tokens"`
} `json:"sentences"`
}
TrainingData represents the structure of the training data JSON.
type WordVectors ¶
SimpleWord2Vec is a basic Word2Vec implementation in Go. does NOT include negative sampling or other important optimizations. WordVectors represents the word embeddings
Source Files