Higher-Order Attribute-Enhancing Heterogeneous Graph Neural Networks
نویسندگان
چکیده
GNNs have been widely used in deep learning on graphs. They learn effective node representations. However, most methods ignore the heterogeneity. Methods designed for heterogeneous graphs, other hand, fail to complex semantic representations because they only use meta-paths instead of meta-graphs. Furthermore, cannot fully capture content-based correlations, as either do not self-attention mechanism or it consider immediate neighbors each node, ignoring higher-order neighbors. We propose a novel Higher-order Attribute-Enhancing (HAE) framework enhancing embedding layer-by-layer manner. Under HAE framework, we GNN (HAE\textsubscript{GNN}) network embeding. HAE\textsubscript{GNN} simultaneously incorporates and meta-graphs rich, semantics, leverages explore nodes' interactions. The unique architecture allows examining first-order well neighborhoods. Moreover, shows good explainability learns importances different is also memory-efficient, avoids per meta-path based matrix calculation. Experimental results show HAE\textsubscript{GNN}'s superior performance against state-of-the-art classification, clustering, visualization, but demonstrate its superiorities terms memory efficiency explainability.
منابع مشابه
Higher Order Recurrent Neural Networks
In this paper, we study novel neural network structures to better model long term dependency in sequential data. We propose to use more memory units to keep track of more preceding states in recurrent neural networks (RNNs), which are all recurrently fed to the hidden layers as feedback through different weighted paths. By extending the popular recurrent structure in RNNs, we provide the models...
متن کاملHigher Order Neural Networks and Neural Networks for Stream Learning
The goal of this thesis is to explore some variations of neural networks. The thesis is mainly split into two parts: a variation of the shaping functions in neural networks and a variation of learning rules in neural networks. In the first part, we mainly investigate polynomial perceptrons a perceptron with a polynomial shaping function instead of a linear one. We prove the polynomial perceptro...
متن کاملTermination Analysis for Higher-Order Attribute Grammars
This paper describes a conservative analysis to detect nontermination in higher-order attribute grammar evaluation caused by the creation of an unbounded number of (finite) trees as local tree-valued attributes, which are then themselves decorated with attributes. This type of non-termination is not detected by a circularity analysis for higherorder attribute grammars. The analysis presented he...
متن کاملHigher-order port-graph rewriting
The biologically inspired framework of port-graphs has been successfully used to specify complex systems. It is the basis of the PORGY modelling tool. To facilitate the specification of proof normalisation procedures via graph rewriting, in this paper we add higher-order features to the original port-graph syntax, along with a generalised notion of graph morphism. We provide a matching algorith...
متن کاملEvolutionary Algorithm Training of Higher Order Neural Networks
This chapter aims to further explore the capabilities of the Higher Order Neural Networks class and especially the Pi-Sigma Neural Networks. The performance of Pi-Sigma Networks is evaluated through several well known neural network training benchmarks. In the experiments reported here, Distributed Evolutionary Algorithms are implemented for Pi-Sigma neural networks training. More specifically,...
متن کاملذخیره در منابع من
با ذخیره ی این منبع در منابع من، دسترسی به آن را برای استفاده های بعدی آسان تر کنید
ژورنال
عنوان ژورنال: IEEE Transactions on Knowledge and Data Engineering
سال: 2021
ISSN: ['1558-2191', '1041-4347', '2326-3865']
DOI: https://doi.org/10.1109/tkde.2021.3074654