Analogical Inference for Multi-relational Embeddings

By Hanxiao Liu, Yuexin Wu, Yiming Yang. 2017

406 citations recorded by the source corpus.

Large-scale multi-relational embedding refers to the task of learning the latent representations for entities and relations in large knowledge graphs. An effective and scalable solution for this problem is crucial for the true success of knowledge-based inference in a broad range of applications. This paper proposes a novel framework for optimizing the latent representations with respect to the \textit{analogical} properties of the embedded entities and relations. By formulating the learning objective in a differentiable fashion, our model enjoys both theoretical power and computational scalability, and significantly outperformed a large number of representative baseline methods on benchmark datasets. Furthermore, the model offers an elegant unification of several well-known methods in multi-relational embedding, which can be proven to be special instantiations of our framework.

Subfields: Artificial Intelligence

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Artificial Intelligence

Citations in the corpus

RippleNet: Propagating User Preferences on the Knowledge Graph for Recommender Systems (2018, background): “Recently, researchers have proposed a great many knowledge graph embedding methods, including translational distance models [3, 14] and semantic matching models [15, 19] (Wewill continue the discussion on KGEmethods in Section 3.”
Representation Learning for Dynamic Graphs: A Survey (2019, methodology): “Liu et al. (2019) also use time as a regularizer, but they turn the representation learning problem into a constrained optimization problem that can be approximated in a reasonable amount of time. As new observations are made, their representations can be updated in a short amount of time and so the”
Knowledge Graph Embedding: A Survey of Approaches and Applications (2017, background): “To model the analogical structures, it further requires relations’ linear maps to be normal and mutually commutative, i.e., normality : MrM > r ¼M>r Mr; 8r 2 IR; commutativity : MrMr0 ¼Mr0Mr; 8r; r0 2 IR: Although ANALOGY represents relations as matrices, these matrices can be simultaneously block-d”
Neural, symbolic and neural-symbolic reasoning on knowledge graphs (2021, methodology): “P Q (Bordes et al., 2013) force h þ r to be close to t TransH (Wang et al., 2014) project entities into the relation-speci ﬁ c hyperplane TransR (Lin et al., 2015) project entities and relations in separate spaces TransD (Ji et al., 2015) determine the project matrices by both entities and relations”
Graph Learning: A Survey (2021, methodology): “Typical models of this type include RESCAL [179], DistMult [180], ANALOGY [181], etc.”
Diachronic Embedding for Temporal Knowledge Graph Completion (2019, unknown): “, [45, 58, 47, 25, 40]).”
TuckER: Tensor Factorization for Knowledge Graph Completion (2019, unknown): “599 ANALOGY (Liu et al., 2017) yes .”
Knowledge Graph Completion: A Review (2020, methodology): “The following are some classical representation methods based on semantic matching: For the representation method based on bilinear model [65], relations after bilinear transformation is used to describe the semantic connection between entity and relation, and to capture various interactions…”
Learning Hierarchy-Aware Knowledge Graph Embeddings for Link Prediction (2019, methodology): “For example, DistMult (Yang et al. 2015) assumes M r to be a diagonal matrix, and ANAL-OGY (Liu, Wu, and Yang 2017) supposes that M r is normal.”
InteractE: Improving Convolution-based Knowledge Graph Embeddings by Increasing Feature Interactions (2019, background): “These score functions can be classiﬁed as translation distance based (Bordes et al. 2013; Xiao, Huang, and Zhu 2016; Wang et al. 2014a) or semantic matching based (Nickel, Rosasco, and Poggio 2016; Liu, Wu, and Yang 2017).”
OpenKE: An Open Toolkit for Knowledge Embedding (2018, background): “…such as linear models (Bordes et al., 2011, 2012, 2014), latent factor models (Sutskever et al., 2009; Jenatton et al., 2012; Yang et al., 2015; Liu et al., 2017), neural models (Socher et al., 2013; Dong et al., 2014), matrix factorization models (Nickel et al., 2011, 2012, 2016; Trouillon et…”
Knowledge Graph Embedding for Link Prediction (2020, methodology): “The included models are: DistMult [71]; ComplEx [61]; ANALOGY [35]; SimplE [27]; HolE [46]; TuckER [6]; TransE [8]; STransE [41]; CrossE [72]; TorusE [13]; RotatE [55]; ConvE [11]; ConvKB [42]; ConvR [25]; CapsE [43]; RSN [19].”
Relational Representation Learning for Dynamic (Knowledge) Graphs: A Survey (2019, methodology): “Other bilinear approaches include HolE [194] whose equivalence to ComplEx has been established [93], and Analogy [148] where the P r matrices are constrained to be block-diagonal.”
A Survey on Knowledge Graphs: Representation, Acquisition, and Applications (2020, methodology): “ANALOGY [22] provides a uniﬁed view of several representative models, including DistMult, ComplEx, and HolE.”
A Survey of Knowledge Graph Reasoning on Graph Types: Static, Dynamic, and Multi-Modal (2022, methodology): “Besides, Analogy [49] makes use of analogical structures by setting speciﬁc constraints, and CrossE [50] considers crossover interactions between entities via a relation-speciﬁc interaction matrix, which both enhance the interactions between entities.”
Adaptive Convolution for Multi-Relational Learning (2019, background): “, 2016), ANALOGY (Liu et al., 2017), TorusE (Ebisu and Ichise, 2017), Gaifman”
SimplE Embedding for Link Prediction in Knowledge Graphs (2018, methodology): “…could be improved or may help improve relational learning in several ways including: 1-building ensembles of SimplE models as [18] do it for DistMult, 2-adding SimplE to the relation-level ensembles of [44], 3-explicitly modelling the analogical structures of relations as in [23], 4-using [8]’s 1-”