Machine Learning Summer School 2020

Today starts Machine Learning Summer School 2020 with great list of speakers on various topics of machine learning, including geometric deep learning. The lectures will be live-streamed on YouTube and are open to everyone.

Manually-curated List of Combinatorial Conferences

Mostly mathematical, with some occasions on CS, here is a manually-curated list of upcoming conferences, workshops, symposiums on combinatorics, among which you can find graph-related topics.

Fresh picks from ArXiv
This week we have papers on theory of GNN, their applications to recommendation and other fields, a historical reference on available graph repositories, and a discussion of peer review vs biblio metrics to assess scientific performance 👨‍⚖️

GNN
• Characterizing the Expressive Power of Invariant and Equivariant Graph Neural Networks
• Building powerful and equivariant graph neural networks with message-passing with Andreas Loukas
• Fast Learning of Graph Neural Networks with Guaranteed Generalizability: One-hidden-layer Case ICML 2020

Applications
• Graph Convolutional Network for Recommendation with Low-pass Collaborative Filters
• Scalable Deep Generative Modeling for Sparse Graphs with Yujia Li, ICML 2020
• GPT-GNN: Generative Pre-Training of Graph Neural Networks KDD 2020
• Molecule Edit Graph Attention Network: Modeling Chemical Reactions as Sequences of Graph Edits

Graph Problems
• Attentional Graph Convolutional Networks for Knowledge Concept Recommendation in MOOCs in a Heterogeneous View with Philip S. Yu
• Bringing Light Into the Dark: A Large-scale Evaluation of Knowledge Graph Embedding Models Under a Unified Framework with Mikhail Galkin
• Graph Policy Network for Transferable Active Learning on Graphs with Jian Tang
• Online Dense Subgraph Discovery via Blurred-Graph Feedback with Masashi Sugiyama, ICML 2020

Surveys
• A survey of repositories in graph theory
• Metrics and peer review agreement at the institutional level

UAI 2020 stats

UAI is a small but strong conference on AI.

Dates: 3-6 Aug
Where: Online
Cost: 125$
Papers available online.

• 580 submissions (vs 450 in 2019)
• 140 accepted (vs 118 in 2019)
• 24.1% acceptance rate (vs 26% in 2019)
• 5 graph papers (4% of total)

Open Problems - Graph Theory and Combinatorics

In addition to Open Problem Garden, there is a list of open problems in graph theory and a corresponding old archive. Sometimes proof to these is just a specific graph that even people without background may find.

Graph Machine Learning research groups: Stephan Günnemann

I do a series of posts on the groups in graph research, previous post is here. The nineth is Stephan Günnemann. His research interests include adversarial attacks on graphs and graph generative models.


Stephan Günnemann (~1984)
- Affiliation: Technical University of Munich
- Education: Ph.D. at RWTH Aachen University in 2012 (supervised by Thomas Seidl);
- h-index: 30;
- Awards: best paper at KDD;
- Interests: graph adversarial attacks; clustering; graph generative models

Graphs with the same degree distribution

Degree distribution plays a key distinctive role between graphs. In networks science there are specific models that generate you a graph according to some distribution of degrees. For example, scale-free networks are the ones with power law degree distribution, which we observe in real world (e.g. social networks). Scale-free networks use preferential attachment mechanism that mimics the way people connect with others in a new society: we connect to people with high degree and people that we know. The Barabási–Albert model is the most famous example of such a model.

What's interesting in some cases is to provide explicitly the degrees that you expect to have in a graph and then generate a graph with this sequence of degrees. There is a model for that too: it's called Chung-Lu model. Yet, in some other cases, you want to generate a graph exactly with some degree sequence. This is quite simple, you just connect pairs of vertices one by one, until you make a desired degree sequence. It shows how many actually there are different graphs with the same degree sequence. Here is an explanation of this.

Beyond Weisfeiler-Lehman: using substructures for provably expressive graph neural networks

This is the second post by Michael Bronstein, where he discussed his recent architecture of GNN. In one sentence, they append information about graph statistics, such as number of 4-cliques, to message-passing mechanism and show that it is theoretically equivalent to k-WL, with fraction of its cost.

For more than 6 months, I wondered why do we try to design GNN that can solve graph isomorphism (GI), if in all cases we are at most as good as already known algorithms to GI. What if we just take a automorphism group of a graph and then append this information to GNN, hoping it will help for downstream tasks. This way we solve GI by default by using automorphism group, and just measure effectiveness of the GNN for the tasks that matter.

Fresh picks from ArXiv
This week highlights clustering with GNN, scalable GNN, recommendation with graphs, and surveys on mathematical perspective of ML 💭

GNN
• Graph Clustering with Graph Neural Networks with Anton Tsitsulin and Bryan Perozzi
• Scaling Graph Neural Networks with Approximate PageRank with Bryan Perozzi, Stephan Günnemann, KDD 2020
• Simple and Deep Graph Convolutional Networks
• AM-GCN: Adaptive Multi-channel Graph Convolutional Networks KDD 2020
• Adaptive Graph Encoder for Attributed Graph Embedding KDD 2020
• A Novel Higher-order Weisfeiler-Lehman Graph Convolution
• Hierarchical Graph Matching Network for Graph Similarity Computation

Applications
• Disentangled Graph Collaborative Filtering SIGIR 2020
• Scene Graph Reasoning for Visual Question Answering with Stephan Günnemann
• An Efficient Neighborhood-based Interaction Model for Recommendation on Heterogeneous Graph with Alexander J. Smola, KDD 2020
• Interactive Path Reasoning on Graph for Conversational Recommendation KDD 2020
• New Hardness Results for Planar Graph Problems in P and an Algorithm for Sparsest Cut

Survey
Mathematical Perspective of Machine Learning
Model-based Reinforcement Learning: A Survey
Boosting Deep Neural Networks with Geometrical Prior Knowledge: A Survey

Pytorch-Geometric version 1.6

In a new release PyG features support for static graph, a zoo of new models, and supplementary frameworks such as DeepSnap.

Graphs and Networks Workshop

There is one-day free online workshop for those who love network science, happening this Friday, July 10.

ICML GRL Workshop Papers

There are 73 interesting short papers on various topics of GML at ICML GRL workshop.

Easiest unsolved case of reconstruction conjecture

We spent some time thinking about reconstruction conjecture and got into a conclusion that no one yet solved a simple case as follows.

You have a graph composed of two parts X and Y. In X you have n vertices, in Y you have 2 vertices only. Vertices in Y are connected to X such that all vertices in X have degree 5 and all vertices in Y have degree 4.

So final graph has only two values of degrees, 4 and 5. It's know that when a graph is regular, it can be reconstructed. Here there are only 2 vertices of different degree, nonetheless the problem becomes harder. In more general case you have 2 vertices of degree a and n vertices of degree a+1, but it seems to be much harder to reason about.

Knowledge Graphs at ACL 2020

Another brilliant post by Michael Galkin on usage of knowledge graphs in NLP at ACL 2020.

"Knowledge graphs demonstrate better capabilities to reveal higher-order interdependencies in otherwise unstructured data."

Content:
1. Question Answering over Structured Data
2. KG Embeddings: Hyperbolic and Hyper-relational
3. Data-to-text NLG: Prepare your Transformer
4. Conversational AI: Improving Goal-Oriented Bots
5. Information Extraction: OpenIE and Link Prediction

Fresh picks from ArXiv
While everyone is at ICML, there is a new batch of interesting fresh papers on GML: new SOTA on graph matching networks, speed-up for GPU computations of GNN, and applications to CV.

GNN
• Inverse Graph Identification: Can We Identify Node Labels Given Graph Labels?
• Hierarchical Graph Matching Networks for Deep Graph Similarity Learning
• NASGEM: Neural Architecture Search via Graph Embedding Method
• Graph Convolutional Networks for Graphs Containing Missing Features
• GE-SpMM: General-purpose Sparse Matrix-Matrix Multiplication on GPUs for Graph Neural Networks

Applications
• Generative Graph Perturbations for Scene Graph Prediction ICML OOL workshop
• Sparse Graph to Sequence Learning for Vision Conditioned Long Textual Sequence Generation ICML workshop
• Graph Factorization Machines for Cross-Domain Recommendation
• Spatio-Temporal Scene Graphs for Video Dialog
• A Generative Graph Method to Solve the Travelling Salesman Problem
• Neural Subgraph Matching with Jure Leskovec
• ASGN: An Active Semi-supervised Graph Neural Network for Molecular Property Prediction


Survey
• A Tutorial on Graph Theory for Brain Signal Analysis

Beyond Weisfeiler-Lehman: approximate isomorphisms and metric embeddings

Last post in a series of posts by M. Bronstein about how we can reformulate the framework of designing GNNs. Most popular approach currently is to show that GNN can be equivalent to WL algorithm, which in turn implies that the algorithm can detect isomorphism of most graphs. However, this is not very valuable in practice, as there are very few graphs. Instead, it would be great to have GNN that preserve some kind of distance (e.g. edit distance) between the graphs in euclidean space, up to some error. This can be seen as a generalization of the current framework, when we care only about the case when the two graphs are isomorphic.

Graph Clustering with Graph Neural Networks

This is a guest post by Anton Tsitsulin (@xgfsru) about their new work called “Graph Clustering with Graph Neural Networks”.

We are proposing a new perspective on unsupervised training of Graph Neural Networks (GNNs) through the lens of graph clustering. Attributed graph clustering is a popular task that poses a unique challenge: we have to balance the information we get from the graph itself and node attributes (features). Can we create an end-to-end GNN model with automatic balance of graph & feature structure?

Intuitively, graph clustering with graph neural networks should be connected with graph pooling: graph clusters that are close in terms of their features can be safely merged together. Not all pooling methods are useful for clustering, however, as some of them do not collapse together nodes (e.g. top-k pooling). State-of-the-art there is MinCut pooling that appeared in ICML’20.

We present DMoN, a GNN loss function that directly optimizes graph modularity in its spectral formulation. Modularity is known to perform better than cut-based metrics in real-world graphs, because of that we obtain an average of 30% better label correlation than the best other neural method, and 45% better scores for graph clustering across 7 datasets.

To better investigate the performance of GNNs for clustering, we also propose a set of synthetic benchmarks. For example, in the attached image we show how the performance of different methods changes with varying signal strength in either graph or node features. We hope that this methodology will allow more thorough evaluation of different models in the future.

TL;DR:
- Graph pooling is very similar to graph clustering, it’s a good idea to evaluate pooling methods on clustering tasks.
- We show how to do clustering in an end-to-end way with GNNs via spectral modularity optimization.
- DMoN offers substantial (30-50% across different metrics) quality increases on 7 datasets.

Paper: https://arxiv.org/abs/2006.16904
Code will be available here soon.