Abstract:<p>In this work, we propose a novel Knowledge Graph Embedding (KGE) strategy, called MöbiusE, in which the entities and relations are embedded to the surface of a Möbius ring. The proposition of such a strategy is inspired by the classic TorusE, in which the addition of two arbitrary elements is subject to a modulus operation. In this sense, TorusE naturally guarantees the critical boundedness of embedding vectors in KGE. However, the nonlinear property of addition operation on Torus ring is uniquely derived by the modulus operation, which in some extent restricts the expressiveness of TorusE. As a further generalization of TorusE, MöbiusE also uses modulus operation to preserve the closeness of addition on it, but the coordinates on Möbius ring interacts with each other in the following way: any vector attaches to the surface of a Mobius ring becomes its opposite one if it moves along its parametric trace by a cycle. Hence, MöbiusE assumes much more nonlinear representativeness than that of TorusE, and in turn it generates much more precise embedding results. In our experiments, MöbiusE outperforms TorusE and other classic embedding strategies in several key indicators.</p>

What problem does this paper attempt to address?

The paper attempts to address the problem of improving the accuracy and expressive power of embeddings in the Knowledge Graph Embedding (KGE) task. Specifically, the authors propose a new knowledge graph embedding strategy called MöbiusE, which embeds entities and relations onto the surface of a Möbius strip. Compared to existing embedding methods based on the Torus ring (such as TorusE), MöbiusE leverages the nonlinear properties of the Möbius strip to generate more accurate embedding results and outperforms other classical embedding strategies in experiments. ### Main Contributions: 1. **Enhanced Flexibility**: MöbiusE provides more flexibility because each point on the Möbius strip has two different representations. 2. **Stronger Expressive Power**: Due to the complex distance function in the Möbius strip, MöbiusE's expressive power far exceeds that of TorusE. 3. **Automatic Truncation**: Like TorusE, MöbiusE can automatically constrain training vectors within the bounded space. 4. **Inheritance of TorusE's Advantages**: MöbiusE inherits all the advantages of TorusE, such as the ability to model symmetry, antisymmetry, inverse operations, and composition operations. ### Problems Addressed: - **Limitations of Existing Methods**: Existing KGE methods, such as TorusE, although performing well in some aspects, have limited expressive power due to their nonlinear properties. MöbiusE addresses this limitation by introducing the complex structure of the Möbius strip. - **Improved Embedding Accuracy**: MöbiusE outperforms existing KGE methods on multiple key metrics, particularly in link prediction tasks, where it can more accurately predict missing triples. ### Experimental Results: - Experiments conducted on the FB15K and WN18 datasets show that MöbiusE outperforms other classical embedding strategies on multiple key metrics (such as MRR, MR, HIT@10, HIT@3, HIT@1). In summary, by introducing the Möbius strip as the embedding space, this paper proposes a new KGE strategy, MöbiusE, which significantly improves the accuracy and expressive power of embeddings.

MöbiusE: Knowledge Graph Embedding on Möbius Ring

ModulE: Module Embedding for Knowledge Graphs

TorusE: Knowledge Graph Embedding on a Lie Group

Meta-Knowledge Transfer for Inductive Knowledge Graph Embedding

Knowledge Graph Embedding with Diversity of Structures

Geometry Interaction Knowledge Graph Embeddings

Knowledge Graph Embedding with 3D Compound Geometric Transformations

Cycle or Minkowski

Knowledge Graph Embedding via Triplet Component Interactions

Möbius–deBruijn: The product of Möbius cube and deBruijn digraph

CoPE: Composition-based Poincaré embeddings for link prediction in knowledge graphs

Knowledge Graph Embedding: An Overview

CompoundE: Knowledge Graph Embedding with Translation, Rotation and Scaling Compound Operations

Cellular complexes and embeddings into Euclidean spaces: Möbius strip, torus, and projective plane

Inference over Unseen Entities, Relations and Literals on Knowledge Graphs

Embedding Knowledge Graph in Function Spaces

Modeling Noncommutative Composition of Relations for Knowledge Graph Embedding

Learning to Understand: Identifying Interactions via the Möbius Transform

Knowledge Graph Embedding Closed under Composition

A Physical Embedding Model for Knowledge Graphs

ConE: Cone Embeddings for Multi-Hop Reasoning over Knowledge Graphs