Entities and relations are represented in a real valued point -wise euclidean space . The various space for projection are shown in the figure

San Jose State University

The space where entities and relations are represented after analysis. Representation space is also called embedding space. The embedding space should satisfy three conditions differentiability, calculation possibility, and definability of a scoring function. The representation space can be categorized into

 - Point-Wise Space
 - Complex Vector Space
 - Gaussian Distribution
 - Manifold and Group:


Representation Space 

A Survey on Knowledge Graphs: Representation, Acquisition and Applications

Point-Wise Space 

Entities and relations are represented in a complex space, $h, t, r \in \mathcal{C}^d$

Complex Vector Space

Uncertainties in entity and relations are modeled using gaussian distribution. Entities and relations are embedded into multi-dimensional gaussian distribution function ,$\mathcal{H} \sim \mathcal{N}(\mu_h,\Sigma_h)  and  \mathcal{T}(\mu_t,\Sigma_t)$

Mean vector u represents entities and relations position.

$\Sigma$ represents entities and relation’s (un)certainties.


Gaussian Distribution

Representation of knowledge in manifold space. Sphere as a manifold embedding is shown in figure.


Learn Before

Related