Graph Neural Networks are Inherently Good Generalizers: Insights by Bridging GNNs and MLPs

Graph neural networks (GNNs), as the de-facto model class for representation learning on graphs, are built upon the multi-layer perceptrons (MLP) architecture with additional message passing layers to allow features to flow across nodes. While conventional wisdom largely attributes the success of GNNs to their advanced expressivity for learning desired functions on nodes' ego-graphs, we conjecture that this is \emph{not} the main cause of GNNs' superiority in node prediction tasks. This paper pinpoints the major source of GNNs' performance gain to their intrinsic generalization capabilities, by introducing an intermediate model class dubbed as P(ropagational)MLP, which is identical to standard MLP in training, and then adopt GNN's architecture in testing. Intriguingly, we observe that PMLPs consistently perform on par with (or even exceed) their GNN counterparts across ten benchmarks and different experimental settings, despite the fact that PMLPs share the same (trained) weights with poorly-performed MLP. This critical finding opens a door to a brand new perspective for understanding the power of GNNs, and allow bridging GNNs and MLPs for dissecting their generalization behaviors. As an initial step to analyze PMLP, we show its essential difference with MLP at infinite-width limit lies in the NTK feature map in the post-training stage. Moreover, though MLP and PMLP cannot extrapolate non-linear functions for extreme OOD data, PMLP has more freedom to generalize near the training support.

Localized Contrastive Learning on Graphs

Contrastive learning methods based on InfoNCE loss are popular in node representation learning tasks on graph-structured data. However, its reliance on data augmentation and its quadratic computational complexity might lead to inconsistency and inefficiency problems. To mitigate these limitations, in this paper, we introduce a simple yet effective contrastive model named Localized Graph Contrastive Learning (Local-GCL in short). Local-GCL consists of two key designs: 1) We fabricate the positive examples for each node directly using its first-order neighbors, which frees our method from the reliance on carefully-designed graph augmentations; 2) To improve the efficiency of contrastive learning on graphs, we devise a kernelized contrastive loss, which could be approximately computed in linear time and space complexity with respect to the graph size. We provide theoretical analysis to justify the effectiveness and rationality of the proposed methods. Experiments on various datasets with different scales and properties demonstrate that in spite of its simplicity, Local-GCL achieves quite competitive performance in self-supervised node representation learning tasks on graphs with various scales and properties.

Geometric Knowledge Distillation: Topology Compression for Graph Neural Networks

We study a new paradigm of knowledge transfer that aims at encoding graph topological information into graph neural networks (GNNs) by distilling knowledge from a teacher GNN model trained on a complete graph to a student GNN model operating on a smaller or sparser graph. To this end, we revisit the connection between thermodynamics and the behavior of GNN, based on which we propose Neural Heat Kernel (NHK) to encapsulate the geometric property of the underlying manifold concerning the architecture of GNNs. A fundamental and principled solution is derived by aligning NHKs on teacher and student models, dubbed as Geometric Knowledge Distillation. We develop non- and parametric instantiations and demonstrate their efficacy in various experimental settings for knowledge distillation regarding different types of privileged topological information and teacher-student schemes.

Towards Out-of-Distribution Sequential Event Prediction: A Causal Treatment

The goal of sequential event prediction is to estimate the next event based on a sequence of historical events, with applications to sequential recommendation, user behavior analysis and clinical treatment. In practice, the next-event prediction models are trained with sequential data collected at one time and need to generalize to newly arrived sequences in remote future, which requires models to handle temporal distribution shift from training to testing. In this paper, we first take a data-generating perspective to reveal a negative result that existing approaches with maximum likelihood estimation would fail for distribution shift due to the latent context confounder, i.e., the common cause for the historical events and the next event. Then we devise a new learning objective based on backdoor adjustment and further harness variational inference to make it tractable for sequence learning problems. On top of that, we propose a framework with hierarchical branching structures for learning context-specific representations. Comprehensive experiments on diverse tasks (e.g., sequential recommendation) demonstrate the effectiveness, applicability and scalability of our method with various off-the-shelf models as backbones.

Trading Hard Negatives and True Negatives: A Debiased Contrastive Collaborative Filtering Approach

Collaborative filtering (CF), as a standard method for recommendation with implicit feedback, tackles a semi-supervised learning problem where most interaction data are unobserved. Such a nature makes existing approaches highly rely on mining negatives for providing correct training signals. However, mining proper negatives is not a free lunch, encountering with a tricky trade-off between mining informative hard negatives and avoiding false ones. We devise a new approach named as Hardness-Aware Debiased Contrastive Collaborative Filtering (HDCCF) to resolve the dilemma. It could sufficiently explore hard negatives from two-fold aspects: 1) adaptively sharpening the gradients of harder instances through a set-wise objective, and 2) implicitly leveraging item/user frequency information with a new sampling strategy. To circumvent false negatives, we develop a principled approach to improve the reliability of negative instances and prove that the objective is an unbiased estimation of sampling from the true negative distribution. Extensive experiments demonstrate the superiority of the proposed model over existing CF models and hard negative mining methods.

Handling Distribution Shifts on Graphs: An Invariance Perspective

There is increasing evidence suggesting neural networks' sensitivity to distribution shifts, so that research on out-of-distribution (OOD) generalization comes into the spotlight. Nonetheless, current endeavors mostly focus on Euclidean data, and its formulation for graph-structured data is not clear and remains under-explored, given the two-fold fundamental challenges: 1) the inter-connection among nodes in one graph, which induces non-IID generation of data points even under the same environment, and 2) the structural information in the input graph, which is also informative for prediction. In this paper, we formulate the OOD problem for node-level prediction on graphs and develop a new domain-invariant learning approach, named Explore-to-Extrapolate Risk Minimization, that facilitates GNNs to leverage invariant graph features for prediction. The key difference to existing invariant models is that we design multiple context explorers (specified as graph editers in our case) that are adversarially trained to maximize the variance of risks from multiple virtual environments. Such a design enables the model to extrapolate from a single observed environment which is the common case for node-level prediction. We prove the validity of our method by theoretically showing its guarantee of a valid OOD solution and further demonstrate its power on various real-world datasets for handling distribution shifts from artificial spurious features, cross-domain transfers and dynamic graph evolution.

Towards Open-World Feature Extrapolation: An Inductive Graph Learning Approach

We target open-world feature extrapolation problem where the feature space of input data goes through expansion and a model trained on partially observed features needs to handle new features in test data without further retraining. The problem is of much significance for dealing with features incrementally collected from different fields. To this end, we propose a new learning paradigm with graph representation and learning. Our framework contains two modules: 1) a backbone network (e.g., feedforward neural nets) as a lower model takes features as input and outputs predicted labels; 2) a graph neural network as an upper model learns to extrapolate embeddings for new features via message passing over a feature-data graph built from observed data. Based on our framework, we design two training strategies, a self-supervised approach and an inductive learning approach, to endow the model with extrapolation ability and alleviate feature-level over-fitting. We also provide theoretical analysis on the generalization error on test data with new features, which dissects the impact of training features and algorithms on generalization performance. Our experiments over several classification datasets and large-scale advertisement click prediction datasets demonstrate that our model can produce effective embeddings for unseen features and significantly outperforms baseline methods that adopt KNN and local aggregation.

From Canonical Correlation Analysis to Self-supervised Graph Neural Networks

We introduce a conceptually simple yet effective model for self-supervised representation learning with graph data. It follows the previous methods that generate two views of an input graph through data augmentation. However, unlike contrastive methods that focus on instance-level discrimination, we optimize an innovative feature-level objective inspired by classical Canonical Correlation Analysis. Compared with other works, our approach requires none of the parameterized mutual information estimator, additional projector, asymmetric structures, and most importantly, negative samples which can be costly. We show that the new objective essentially 1) aims at discarding augmentation-variant information by learning invariant representations, and 2) can prevent degenerated solutions by decorrelating features in different dimensions. Our theoretical analysis further provides an understanding for the new objective which can be equivalently seen as an instantiation of the Information Bottleneck Principle under the self-supervised setting. Despite its simplicity, our method performs competitively on seven public graph datasets.

Inductive Relational Matrix Completion

Data sparsity and cold-start issues emerge as two major bottlenecks for matrix completion in the context of user-item interaction matrix. We propose a novel method that can fundamentally address these issues. The main idea is to partition users into support users, which have many observed interactions (i.e., non-zero entries in the matrix), and query users, which have few observed entries. For support users, we learn their transductive preference embeddings using matrix factorization over their interactions (a relatively dense sub-matrix). For query users, we devise an inductive relational model that learns to estimate the underlying relations between the two groups of users. This allows us to attentively aggregate the preference embeddings of support users in order to compute inductive embeddings for query users. This new method can address the data sparsity issue by generalizing the behavior patterns of warm-start users to others and thus enables the model to also work effectively for cold-start users with no historical interaction. As theoretical insights, we show that a general version of our model does not sacrifice any expressive power on query users compared with transductive matrix factorization under mild conditions. Also, the generalization error on query users is bounded by the numbers of support users and query users' observed interactions. Moreover, extensive experiments on real-world datasets demonstrate that our model outperforms several state-of-the-art methods by achieving significant improvements on MAE and AUC for warm-start, few-shot (sparsity) and zero-shot (cold-start) recommendation.

Learning Latent Process from High-Dimensional Event Sequences via Efficient Sampling

We target modeling latent dynamics in high-dimension marked event sequences without any prior knowledge about marker relations. Such problem has been rarely studied by previous works which would have fundamental difficulty to handle the arisen challenges: 1) the high-dimensional markers and unknown relation network among them pose intractable obstacles for modeling the latent dynamic process; 2) one observed event sequence may concurrently contain several different chains of interdependent events; 3) it is hard to well define the distance between two high-dimension event sequences. To these ends, in this paper, we propose a seminal adversarial imitation learning framework for high-dimension event sequence generation which could be decomposed into: 1) a latent structural intensity model that estimates the adjacent nodes without explicit networks and learns to capture the temporal dynamics in the latent space of markers over observed sequence; 2) an efficient random walk based generation model that aims at imitating the generation process of high-dimension event sequences from a bottom-up view; 3) a discriminator specified as a seq2seq network optimizing the rewards to help the generator output event sequences as real as possible. Experimental results on both synthetic and real-world datasets demonstrate that the proposed method could effectively detect the hidden network among markers and make decent prediction for future marked events, even when the number of markers scales to million level.