Graph Machine Learning

Fresh picks from ArXiv
This week on ArXiv: PDE-inspired GNNs, proves to the conjectures, and a new benchmark for graph completion 🧵

If I forgot to mention your paper, please shoot me a message and I will update the post.

GNNs
* Distilling Holistic Knowledge with Graph Neural Networks ICCV 2021
* Fully Hyperbolic Graph Convolution Network for Recommendation CIKM 2021
* Jointly Attacking Graph Neural Network and its Explanations
* LEO: Learning Energy-based Models in Graph Optimization
* PDE-GCN: Novel Architectures for Graph Neural Networks Motivated by Partial Differential Equations
* Shift-Robust GNNs: Overcoming the Limitations of Localized Graph Training data with Bryan Perozzi
* AdaGNN: A multi-modal latent representation meta-learner for GNNs based on AdaBoosting

Math
* Isomorphisms between random graphs
* Edge Partitions of Complete Geometric Graphs (Part 1)

Survey
* Are Missing Links Predictable? An Inferential Benchmark for Knowledge Graph Completion
* Influence Maximization in Social Networks: A Survey of Behaviour-Aware Methods

GDL Course

A course that follows closely the geometric deep learning book. It contains 12 lectures, 2 tutorials, and 4 seminars covering topics such as graphs, sets, grids, groups, geodesics, gauges, and time warping. Videos and slides are available.

Essays on Data Science

A great collection of blog posts on machine learning and computer science covering topics such as infinitely wide neural nets, markov models, and graph deep learning.

Knowledge Graphs in Natural Language Processing @ ACL 2021

A regular update from Michael Galkin on the SOTA applications of KG in the world of words:

Neural Databases & Retrieval
KG-augmented Language Models
KG Embeddings & Link Prediction
Entity Alignment
KG Construction, Entity Linking, Relation Extraction
KGQA: Temporal, Conversational, and AMR.

Graph Neural Networks: Algorithms and Applications

A great presentation by Jian Tang about GNN basics, training many layers, self-supervised learning and statistical relational learning.

Foundations of Graph Neural Networks Course

A new upcoming course by Zak Jost (you may remember his videos on GNNs) on the foundations of GNN which covers such topics as
- Neural Message Passing
- Fourier Transforms, Graph Wavelets and Spectral Convolutions
- Permutation Symmetries
- Representational capacity of GNNs
- Graph fundamentals like the Laplacian and graph isomorphism.

Fresh picks from ArXiv
This week on ArXiv: time series recovery, GNN challenge winning solutions, and benchmark for scene graph generation 🌳

If I forgot to mention your paper, please shoot me a message and I will update the post.

Applications
* Multivariate Time Series Imputation by Graph Neural Networks
* Temporal-Relational Hypergraph Tri-Attention Networks for Stock Trend Prediction
* Graph Constrained Data Representation Learning for Human Motion Segmentation
* The Graph Neural Networking Challenge: A Worldwide Competition for Education in AI/ML for Networks
* Image Scene Graph Generation (SGG) Benchmark
* Structack: Structure-based Adversarial Attacks on Graph Neural Networks

GNN User Group Videos

Videos from the last Thursday meeting of GNN user group are available now. This includes updates of DGL library, storing node feature for large graphs, and locally private GNNs.

Graph Machine Learning research groups: Shuiwang Ji

I do a series of posts on the groups in graph research, previous post is here. The 32nd is Shuiwang Ji, a professor at Texas A&M University. His teams were awarded at OGB-LSC and AI Cures challenges. He also recently advised graph libraries such as MoleculeX and DIG.

Shuiwang Ji (~1982)
- Affiliation: Texas A&M University
- Education: Ph.D. at Arizona State University in 2008 (advisor: Jieping Ye)
- h-index 44
- Interests: GNNs, self-supervised learning, surveys, libraries.
- Awards: best papers at KDD, WWW, ACM Distinguished Member

Graph Convolutional Neural Networks to Analyze Complex Carbohydrates

A blog post by Daniel Bojar about an application of GNN to analyzing glycan sequences and their proposed GNN architecture called SweetNet. There are other coverages of this work (here and here). The paper is here and the code is here.

Header-Only C++ Library for Graph Representation and Algorithms

In case you need the speed of C++ for the well-known graph algorithms there is a nice repo that collects many of them.

Graph Neural Networks User Group: July meeting

This month GNN user group talks about a new release of DGL and applications of GNNs. Please join this Thursday!

4:00 - 4:15 PM (PDT): DGL 0.7 release(Dr. Minjie Wang, Amazon)
4:15 - 4:30 PM (PDT): Storing Node Features in GPU memory to speedup billion-scale GNN training (Dr. Dominique LaSalle, NVIDIA)
4:30 - 5:00 PM (PDT): Locally Private Graph Neural Networks (Sina Sajadmanesh, Idiap Research Institute, Switzerland).
5:00 - 5:30 PM (PDT): Graph Embedding and Application in Meituan (Mengdi Zhang, Meituan).

Fresh picks from ArXiv
This week on ArXiv: SOTA for protein energy prediction, another solution to OGB-LSC challenge, and a new dataset based on Wikipedia 📚

If I forgot to mention your paper, please shoot me a message and I will update the post.

Applications
* X-GGM: Graph Generative Modeling for Out-of-Distribution Generalization in Visual Question Answering
* Structure-aware Interactive Graph Neural Networks for the Prediction of Protein-Ligand Binding Affinity KDD 2021

GNNs
* Local2Global: Scaling global representation learning on graphs via local training
* Ego-GNNs: Exploiting Ego Structures in Graph Neural Networks with William L. Hamilton
* Bridging the Gap between Spatial and Spectral Domains: A Theoretical Framework for Graph Neural Networks
* Large-scale graph representation learning with very deep GNNs and self-supervision with Petar Veličković
* Group Contrastive Self-Supervised Learning on Graphs with Shuiwang Ji

Datasets
* WikiGraphs: A Wikipedia Text - Knowledge Graph Paired Dataset with Oriol Vinyals

Interpretable Deep Learning for New Physics Discovery

In this video, Miles Cranmer (Princeton) discusses a method for converting a neural network into an analytic equation using a particular set of inductive biases. The technique relies on a sparsification of latent spaces in a deep neural network, followed by symbolic regression. In their paper, they demonstrate that they can recover physical laws for various simple and complex systems. For example, they discover gravity along with planetary masses from data; they learn a technique for doing cosmology with cosmic voids and dark matter halos; and they show how to extract the Euler equation from a graph neural network trained on turbulence data.

labml.ai Annotated PyTorch Paper Implementations

A very cool collection of popular deep learning blocks, nicely formatted in the browser with extensive comments. Among others there is a GAT implementation.

Awesome Explainable Graph Reasoning

An awesome collection of research papers and software related to explainability in graph machine learning, provided by AstraZeneca. It covers papers on explainable predictions and reasoning, libraries, and survey papers.

Graph Papers at ICML 2021

ICML 2021 is happening this week and here is a list of all relevant graph papers that you can encounter there. There are papers on improving expressiveness, explainability, robustness, normalization and theory.

Fresh picks from ArXiv
This week on ArXiv: QA in images, graph matching, and learning robot dynamics 🤖

If I forgot to mention your paper, please shoot me a message and I will update the post.

GNNs
* Train on Small, Play the Large: Scaling Up Board Games with AlphaZero and GNN
* Reasoning-Modulated Representations with Petar Veličković and Thomas Kipf
* SENSORIMOTOR GRAPH: Action-Conditioned Graph Neural Network for Learning Robotic Soft Hand Dynamics
* Scalable Optimal Transport in High Dimensions for Graph Distances, Embedding Alignment, and More with Stephan Günnemann
* Graphhopper: Multi-Hop Scene Graph Reasoning for Visual Question Answering with Stephan Günnemann
* Elastic Graph Neural Networks

Survey
* A Survey of Knowledge Graph Embedding and Their Applications

Effortless Distributed Training of Ultra-Wide GCNs

A great post about distributed training of GNNs on large graphs. The architecture splits the GNN into several submodules where each is trained independently on separate GPUs, providing the flexibility to increase significantly the hidden dimension of embeddings. As such this approach is GCN model agnostic, compatible with existing sampling methods, and performs the best in very large graphs.

LOGML Videos

LOGML is an exciting summer school with projects and talks about graph ML happening this week. A collection of videos that includes presentations of the cutting edge research as well as industrial applications from leading companies are available now for everyone.