Open-World Semi-Supervised Learning for Node Classification (2024-03-18T00:00:00.000000Z)

TL;DR

This paper proposes an IMbalance-A ware method named OpenIMA for Open-world semi-supervised node classification, which trains the node classification model from scratch via contrastive learning with bias-reduced pseudo labels.

Abstract

Open-world semi-supervised learning (Open-world SSL) for node classification, that classifies unlabeled nodes into seen classes or multiple novel classes, is a practical but under-explored problem in the graph community. As only seen classes have human labels, they are usually better learned than novel classes, and thus exhibit smaller intra-class variances within the embedding space (named as imbalance of intra-class variances between seen and novel classes). Based on empirical and theoretical analysis, we find the variance imbalance can negatively impact the model performance. Pre-trained feature encoders can alleviate this issue via producing compact representations for novel classes. However, creating general pre-trained encoders for various types of graph data has been proven to be challenging. As such, there is a demand for an effective method that does not rely on pre-trained graph encoders. In this paper, we propose an IMbalance-A ware method named OpenIMA for Open-world semi-supervised node classification, which trains the node classification model from scratch via contrastive learning with bias-reduced pseudo labels. Extensive experiments on seven popular graph benchmarks demonstrate the effectiveness of OpenIMA, and the source code has been available on GitHub11https://github.com/RUCKBReasoning/OpenIMA.

Authors

Jing Zhang

2 papers

Cuiping Li

2 papers

Yanling Wang

1 papers

TL;DR

Abstract

Authors

References70 items

Contrastive Pseudo Learning for Open-World DeepFake Attribution

G2Pxy: Generative Open-Set Node Classification on Graphs with Proxy Unknowns

Open-World Graph Active Learning for Node Classification

All in One: Multi-Task Prompting for Graph Neural Networks

Learning Semi-supervised Gaussian Mixture Models for Generalized Category Discovery

Imbalanced Node Classification Beyond Homophilic Assumption

GraphPrompt: Unifying Pre-Training and Downstream Tasks for Graph Neural Networks

Zero-Knowledge Zero-Shot Learning for Novel Visual Category Discovery

PromptCAL: Contrastive Affinity Learning via Auxiliary Prompts for Generalized Novel Category Discovery

Parametric Classification for Generalized Category Discovery: A Baseline Study

A Dynamic Variational Framework for Open-World Node Classification in Structured Sequences

Learning on Graphs with Out-of-Distribution Nodes

GPPT: Graph Pre-training and Prompt Tuning to Generalize Graph Neural Networks

OpenCon: Open-world Contrastive Learning

OpenLDN: Learning to Discover Novel Classes for Open-World Semi-Supervised Learning

LaSSL: Label-Guided Self-Training for Semi-supervised Learning

Not All Parameters Should Be Treated Equally: Deep Safe Semi-supervised Learning under Class Distribution Mismatch

ClusterSCL: Cluster-Aware Supervised Contrastive Learning on Graphs

How to Exploit Hyperspherical Embeddings for Out-of-Distribution Detection?

Class-Aware Contrastive Semi-Supervised Learning

Generalized Category Discovery

Lifelong learning on evolving graphs under the constraints of imbalanced classes and new classes

OpenWGL: open-world graph learning for unseen class node classification

Decoupling Representation Learning and Classification for GNN-based Anomaly Detection

ImGAGN: Imbalanced Network Embedding via Generative Adversarial Graph Networks

OpenMatch: Open-set Consistency Regularization for Semi-supervised Learning with Outliers

Semi-Supervised Contrastive Learning With Similarity Co-Calibration

MiCE: Mixture of Contrastive Experts for Unsupervised Image Clustering

SimCSE: Simple Contrastive Learning of Sentence Embeddings

Open-World Semi-Supervised Learning

An Image is Worth 16x16 Words: Transformers for Image Recognition at Scale

Safe Deep Semi-Supervised Learning for Unseen-Class Unlabeled Data

Unsupervised Learning of Visual Features by Contrasting Cluster Assignments

GCC: Graph Contrastive Coding for Graph Neural Network Pre-Training

Prototypical Contrastive Learning of Unsupervised Representations

Open Graph Benchmark: Datasets for Machine Learning on Graphs

Supervised Contrastive Learning

OpenMix: Reviving Known Knowledge for Discovering Novel Visual Categories in an Open World

Clustering based Contrastive Learning for Improving Face Representations

Semi-Supervised Learning under Class Distribution Mismatch

Conditional Gaussian Distribution Learning for Open Set Recognition

Automatically Discovering and Learning New Visual Categories with Ranking Statistics

Multi-Stage Self-Supervised Learning for Graph Convolutional Networks

Recent Advances in Open Set Recognition: A Survey

Pitfalls of Graph Neural Network Evaluation

How Powerful are Graph Neural Networks?

Representation Learning with Contrastive Predictive Coding

GeniePath: Graph Neural Networks with Adaptive Receptive Paths

Learning to cluster in order to Transfer across domains and tasks

Graph Attention Networks

Inductive Representation Learning on Large Graphs

Zero-Shot Learning — The Good, the Bad and the Ugly

Semi-Supervised Classification with Graph Convolutional Networks

An Empirical Study and Analysis of Generalized Zero-Shot Learning for Object Recognition in the Wild

Deep Residual Learning for Image Recognition

Toward Open Set Recognition

Speeding up K-Means Algorithm by GPUs

Efficiently learning mixtures of two Gaussians

Web-scale k-means clustering

Parallel K-Means Clustering Based on MapReduce

k-means++: the advantages of careful seeding

Silhouettes: a graphical aid to the interpretation and validation of cluster analysis

The Hungarian method for the assignment problem

RIGNN: A Rationale Perspective for Semi-supervised Open-world Graph Classification

Mutual Information-based Generalized Category Discovery

Lifelong Learning in Evolving Graphs with Limited Labeled Data and Unseen Class Detection

Do Transformers Really Perform Badly for Graph Representation?

Generalized Zero-Shot Learning with Deep Calibration Network

Dropout: a simple way to prevent neural networks from overfitting

license agreement with IEEE. Restrictions apply

Field of Study

Journal Information

Name

Page

Venue Information

Name