Graphs or networks are ubiquitous structures that appear in a multitude of complex systems like social networks, biological networks, knowledge graphs, world wide web, transportation networks, and many more. Real-world networks are massive and unstructured, apart from dynamic and multi-modal. Many existing domains can benefit from data analysis modelled as a networks problem that provide many computational and algorithmic challenges. Essentially, networks provide enormous potential to address long-standing scientific questions and particularly inform the design of several machine learning applications. Graph-based learning and reasoning approaches offer a way to integrate symbolic reasoning (which offer more interpretability) with the representation learning capabilities of deep neural networks to introduce causality, interpretability, and transferability.

The third year of Machine Learning for Graphs special session aims to bring researchers across disciplines to share their innovative ideas on machine learning for graphs and leverage existing methodologies across several application domains. This special session will also serve as a common ground to showcase recent advancements in ML for graphs, build collaborations across disciplines, share benchmark datasets for graph-based ML algorithm evaluation, and inspire machine learning for graphs research in domains where there are limitations in the existing approaches. Authors of the best papers from this special session will have an opportunity to extend their work and publish in selected journals.

Scope and Topics:

We welcome novel research papers on the following algorithms and applications, including but not limited to:

Algorithms
Graph representation learning
Hyperbolic graph embedding
ML on Signed networks
ML on multi-layer, multi-modal, and heterogeneous graphs
ML on knowledge graphs
ML on evolving graphs and graph streams
ML on cascades and cascade growth
ML on low-resource settings
ML on Test-Time Generalization
Network growth models ● Graph summarization
Graph partitioning
Graph matching
Graph generative models
Network fusion
Graph reinforcement learning
Scalable ML algorithms for graphs

Applicatioins in computational social science
Social network analysis
Cyberbullying
Affective polarization
Echo chambers
Civil unrest
Fake news and misinformation spread
Hate speech
Population migration
Local and global politics

Applications in Computer Vision, Natural Language Processing and Speech Processing
Question Answering using Knowledge Graphs and Deep Learning
Scene graph generation
Activity understanding from multimodal data
Image and Video captioning
Knowledge graphs for multimodal understanding
Neural-symbolic integration
Explainable methods for visual understanding
Common sense knowledge graph construction
Applying knowledge graph embeddings to real world scenarios
Speaker Diarization, Speech Emotion Recognition and Speech Enhancement

Applications in Health and Medicine
Health informatics and analytics
Health misinformation
Disease epidemics
Genomics
Population health
Synthetic population
Drug discovery