embeddingTrainer: Poincare-embedding trainer

In hwchang1201/poincare.embeddings: R-package implementation of Poincaré Embeddings for Learning Hierarchical Representations.

Description

This function performs optimization with respect to the objective function. Please refer to 'vignette' to find the equation for the objective function and detailed ideas(e.g. how to optimize).

Usage

embeddingTrainer(POS, NEG, entity, theta_dim = 2, N_epoch = 100,
  lr = 0.001, n_neg = 4)

Arguments

POS	- A 2-row positive relation matrix that contains positive relation entities columnwisely.
NEG	- A 2-row negative relation matrix that contains negative relation entities columnwisely.
entity	- A vector of all the entities in the tree-shape dataset.
theta_dim	- The dimension of the embedding space.
N_epoch	- The number of epochs.
lr	- The learning rate.
n_neg	- The number of negative samples for each iteration.

Value

A trained matrix theta.

Examples

statistics_yaml <- yaml::yaml.load(statistics)
statistics_tree <- data.tree::as.Node(statistics_yaml)
dataset <- dataLoader(statistics_tree)
embeddingTrainer(dataset$POS, dataset$NEG, dataset$entity, 2, 100, 0.001, 5)

hwchang1201/poincare.embeddings documentation built on Dec. 9, 2019, 1:30 p.m.