Timezone: »

Graph Representation Learning
Will Hamilton · Rianne van den Berg · Michael Bronstein · Stefanie Jegelka · Thomas Kipf · Jure Leskovec · Renjie Liao · Yizhou Sun · Petar Veličković

Fri Dec 13 08:00 AM -- 06:00 PM (PST) @ West Exhibition Hall A
Event URL: https://grlearning.github.io »

Graph-structured data is ubiquitous throughout the natural and social sciences, from telecommunication networks to quantum chemistry. Building relational inductive biases into deep learning architectures is crucial if we want systems that can learn, reason, and generalize from this kind of data. Furthermore, graphs can be seen as a natural generalization of simpler kinds of structured data (such as images), and therefore, they represent a natural avenue for the next breakthroughs in machine learning.

Recent years have seen a surge in research on graph representation learning, including techniques for deep graph embeddings, generalizations of convolutional neural networks to graph-structured data, and neural message-passing approaches inspired by belief propagation. These advances in graph neural networks and related techniques have led to new state-of-the-art results in numerous domains, including chemical synthesis, 3D-vision, recommender systems, question answering, and social network analysis.

The workshop will consist of contributed talks, contributed posters, and invited talks on a wide variety of methods and problems related to graph representation learning. We will welcome 4-page original research papers on work that has not previously been published in a machine learning conference or workshop. In addition to traditional research paper submissions, we will also welcome 1-page submissions describing open problems and challenges in the domain of graph representation learning. These open problems will be presented as short talks (5-10 minutes) immediately preceding a coffee break to facilitate and spark discussions.

The primary goal for this workshop is to facilitate community building; with hundreds of new researchers beginning projects in this area, we hope to bring them together to consolidate this fast-growing area of graph representation learning into a healthy and vibrant subfield.

Fri 8:40 a.m. - 9:00 a.m.
Opening remarks
Will Hamilton
Fri 9:00 a.m. - 9:30 a.m.

Neural architectures and many learning environments can conveniently be expressed by graphs. Interestingly, it has been recently shown that the notion of receptive field and the correspondent convolutional computation can nicely be extended to graph-based data domains with successful results. On the other hand, graph neural networks (GNN) were introduced by extending the notion of time-unfolding, which ended up into a state-based representation along with a learning process that requires state relaxation to a fixed-point. It turns out that algorithms based on this approach applied to learning tasks on collections of graphs are more computationally expensive than recent graph convolutional nets.

In this talk we advocate the importance of refreshing state-based graph representations in the spirit of the early introduction of GNN for the case of “network domains” that are characterized by a single graph (e.g. traffic nets, social nets). In those cases, data over the graph turn out to be a continuous stream, where time plays a crucial role and blurs the classic statistical distinction between training and test set. When expressing the graphical domain and the neural network within the same Lagrangian framework for dealing with constraints, we show novel learning algorithms that seem to be very appropriate for network domains. Finally, we show that in the proposed learning framework, the Lagrangian multipliers are associated with the delta term of Backpropagation, and provide intriguing arguments on its biological plausibility.

Marco Gori
Fri 9:30 a.m. - 10:00 a.m.

I'll describe a series of studies that use graph networks to reason about and interact with complex physical systems. These models can be used to predict the motion of bodies in particle systems, infer hidden physical properties, control simulated robotic systems, build physical structures, and interpret the symbolic form of the underlying laws that govern physical systems. More generally, this work underlines graph neural networks' role as a first-class member of the deep learning toolkit.

Peter Battaglia
Fri 10:00 a.m. - 10:30 a.m.
Open Challenges - Spotlight Presentations (Spotlight)
Xavier Sumba Toral, Haggai Maron, Arinbjörn Kolbeinsson
Fri 10:30 a.m. - 11:00 a.m.
Coffee Break (Break)
Fri 11:00 a.m. - 11:30 a.m.
Andrew McCallum: Learning DAGs and Trees with Box Embeddings and Hyperbolic Embeddings (Talk)
Andrew McCallum
Fri 11:30 a.m. - 12:30 p.m.
Poster Session #1 (Poster Session)
Adarsh Jamadandi, Sophia Sanborn, Huaxiu Yao, Chen Cai, Yu Chen, Jean-Marc Andreoli, Niklas Stoehr, Shih-Yang Su, Tony Duan, Fabio Ferreira, Davide Belli, Amit Boyarski, Zack Ye, Elahe Ghalebi, Arindam Sarkar, MAHMOUD KHADEMI, Evgeniy Faerman, Joey Bose, Jiaqi Ma, Lin Meng, Seyed Mehran Kazemi, Guangtao Wang, Tong Wu, Yuexin Wu, Chaitanya Joshi, Marc Brockschmidt, Daniele Zambon, Colin Graber, Rafaël Van Belle, Osman Malik, Xavier Glorot, Mario Krenn, Christopher Cameron, Binxuan Huang, George Stoica, Alexia Toumpa
Fri 12:30 p.m. - 1:30 p.m.
Lunch (Break)
Fri 1:30 p.m. - 1:45 p.m.
Outstanding Contribution Talk: Pre-training Graph Neural Networks (Talk)
Bowen Liu
Fri 1:45 p.m. - 2:00 p.m.
Outstanding Contribution Talk: Variational Graph Convolutional Networks (Talk)
Edwin Bonilla
Fri 2:00 p.m. - 2:15 p.m.
Outstanding Contribution Talk: Probabilistic End-to-End Graph-based Semi-Supervised Learning (Talk)
mariana vargas vieyra
Fri 2:15 p.m. - 2:45 p.m.
Wengong Jin: Representation and Synthesis of Molecular Graphs (Talk)
Wengong Jin
Fri 2:45 p.m. - 3:15 p.m.
Presentation and Discussion: Open Graph Benchmark (Talk and Discussion)
Jure Leskovec
Fri 3:15 p.m. - 4:15 p.m.
Poster Session #2 (Poster Session)
Yunzhu Li, Pete Meltzer, Jianing Sun, Guillaume SALHA, Marin Vlastelica Pogančić, Chia-Cheng Liu, Fabrizio Frasca, Marc-Alexandre Côté, Vikas Verma, Abdulkadir CELIKKANAT, Pierluca D'Oro, Priyesh Vijayan, Maria Schuld, Petar Veličković, Kshitij Tayal, Yulong Pei, Hao Xu, Lei Chen, Pengyu Cheng, Ines Chami, Dongkwan Kim, Guilherme Gomes, Lukasz Maziarka, Jessica Hoffmann, Ron Levie, Antonia Gogoglou, Shunwang Gong, Federico Monti, Wenlin Wang, Yan Leng, Salvatore Vivona, Daniel Flam-Shepherd, Chester Holtz, Li Zhang, MAHMOUD KHADEMI, I-Chung Hsieh, Aleksandar Stanić, Ziqiao Meng, Yuhang Jiao
Fri 4:15 p.m. - 4:45 p.m.

Numerous real world applications involve discrete optimization problems on graphs, many of which are NP-hard and hence developing effective combinatorial algorithms is a research challenge. In this talk, I will show how leveraging the power of machine learning, and in particular graph representation learning, can provide a new paradigm for designing data-driven algorithms to solve combinatorial graph optimization problems. Our approaches automatically learn solution strategies from distribution of instances by explicitly considering the combinatorial task during training. We show that we match and often outperform hand-designed algorithms both with learning greedy algorithms for Minimum Vertex Cover, Maxcut and TSP, as well as when using our new framework ClusterNet to learn a graph representation for an efficient differential kmeans proxy for graph problems such as partitioning for Community Detection and node selection for Facility Location.

Bistra Dilkina
Fri 4:45 p.m. - 5:15 p.m.
Marinka Zitnik: Graph Neural Networks for Drug Discovery and Development (Talk)
Marinka Zitnik
Fri 5:15 p.m. - 5:30 p.m.
Invited Presentation: Deep Graph Library (Talk)
Zheng Zhang

Author Information

Will Hamilton (McGill)
Rianne van den Berg (Google Brain)
Michael Bronstein (USI)
Stefanie Jegelka (MIT)
Thomas Kipf (University of Amsterdam)
Jure Leskovec (Stanford University and Pinterest)
Renjie Liao (University of Toronto)
Yizhou Sun (UCLA)
Petar Veličković (DeepMind)

More from the Same Authors