Learning Graph Representation With Generative Adversarial Nets

Abstract

The goal of graph representation learning is to embed each vertex in a graph into a low-dimensional vector space. Existing graph representation learning methods can be classified into two categories: generative models that learn the underlying connectivity distribution in the graph, and discriminative models that predict the probability of edge existence between a pair of vertices. In this paper, we propose Graph-GAN, an innovative graph representation learning framework unifying above two classes of methods, in which the generative model and discriminative model play a game-theoretical minimax game. Specifically, for a given vertex, the generative model tries to fit its underlying true connectivity distribution over all other vertices and produces "fake" samples to fool the discriminative model, while the discriminative model tries to detect whether the sampled vertex is from ground truth or generated by the generative model. With the competition between these two models, both of them can alternately and iteratively boost their performance. Moreover, when considering the implementation of generative model, we propose a novel graph softmax to overcome the limitations of traditional softmax function, which can be proven satisfying desirable properties of normalization, graph structure awareness, and computational efficiency. Through extensive experiments on real-world datasets, we demonstrate that Graph-GAN achieves substantial gains in a variety of applications, including link prediction, node classification, and recommendation, over state-of-the-art baselines.

Graph Adversarial Training: Dynamically Regularizing Based on Graph Structure

“…and detecting fake examples. Recently, several GAN-based models are proposed to learn graph embeddings, which either generate fake nodes and edges to augment embedding learning [39], [40] or smooth the leaned embeddings to follow a prior distribution [41]- [44]. However, using two different networks inevitably doubles the computation of model training and the labor of parameter tuning of GAN-based methods.…”

An end-to-end neighborhood-based interaction model for knowledge-enhanced recommendation

“…Introducing high-order neighborhood information has shown effective [23,25,28] in graph-based recommendation, thus we introduce graph convolution network (GCN) [15] and graph attention network (GAT) [24] to encode high-order neighborhood information for NI model. Graph convolution network computes high-order node representations by stacking several graph convolution layers.…”

Generating Reliable Friends via Adversarial Training to Improve Social Recommendation

“…The main idea of adversarial learning is to simulate a minimax game with the generator attempting to imitate the genuine data distribution while the discriminator aiming to differentiate fake examples from the real data. A few pioneering works [20], [21], [36]- [41] have explored the adversarial learning in recommender systems. IRGAN [20] is the first influential IR model constructed based on GAN.…”

“…Feed the obtained vector into the second Gumbel-Softmax layer; 12 Get a one-hot vector representing the item z; 13 Deliver z to the discriminator; Receive the generated item z from G θ ; 19 Sample a postive item i and a negative item j; 20 Train the BPR model with sampled items; 21 Update D φ based on Eq. 10 and keep G θ fixed; 22 Update G θ based on Eq.…”