Adaptive Convolution for Multi-Relational Learning

Abstract

We consider the problem of learning distributed representations for entities and relations of multi-relational data so as to predict missing links therein. Convolutional neural networks have recently shown their superiority for this problem, bringing increased model expressiveness while remaining parameter efficient. Despite the success, previous convolution designs fail to model full interactions between input entities and relations, which potentially limits the performance of link prediction. In this work we introduce ConvR, an adaptive convolutional network designed to maximize entity-relation interactions in a convolutional fashion. ConvR adaptively constructs convolution filters from relation representations, and applies these filters across entity representations to generate convolutional features. As such, ConvR enables rich interactions between entity and relation representations at diverse regions, and all the convolutional features generated will be able to capture such interactions. We evaluate ConvR on multiple benchmark datasets. Experimental results show that: (1) ConvR performs substantially better than competitive baselines in almost all the metrics and on all the datasets; (2) Compared with stateof-the-art convolutional models, ConvR is not only more effective but also more efficient. It offers a 7% increase in MRR and a 6% increase in Hits@10, while saving 12% in parameter storage.

RAGAT: Relation Aware Graph Attention Network for Knowledge Graph Completion

“…Variants of relation specific parameters θ r RAGAT restricts θ r to be W r for training speed. In fact, θ r can be extended to more parameterized forms like multi-Layer perceptron, convolution kernel [31], etc. More interestingly, we find our idea associated with two most recent works, ParamE [45] and CoPER [46].…”

“…InteractE [29] improves the performance of ConvE by feature permutation, checkered reshaping, and circular convolution. More CNN based approaches include ConvKB [30], ConvR [31], CapsE [32].…”

“…• Neural: Methods that leverage non-linear neural networks like convolutional neural networks and graph neural networks to estimate scores. CNN based method includes ConvE [20], ConvR [31], InteractE [29]. For a fair comparison, we compare against several recent TABLE 2.…”

DAPath: Distance-aware knowledge graph reasoning based on deep reinforcement learning

“…This type of methods are more suitable for the single-hop reasoning. Tensor decomposition models [26], such as DistMult [27], ComplEx [28], Analogy [29], SimplE [30], HolE [31], TuckER [32], and deep learning models, such as ConvE [33], ConvKB [34], ConvR [35], CapsE [36], RSN [37], are also used to represent entities and relations.…”

NeuInfer: Knowledge Inference on N-ary Facts

“…ConvE (Dettmers et al, 2018) models entity inference process via 2D convolution over the reshaped then concatenated embedding of the known entity and relation. ConvR (Jiang et al, 2019) further adaptively constructs convolution filters from relation embedding and applies these filters across entity embedding to generate convolutional features. SENN (Guan et al, 2018) models the inference processes of head entities, tail entities, and relations via fullyconnected neural networks, and integrates them into a unified framework.…”