Interesting Graph and Machine Learning papers

Meta Learning

“MAML- Model-Agnostic Meta-Learning for Fast Adaptation of Deep Networks”
Finn et al. - ICML 2017.

“Reptile-On First-Order Meta-Learning Algorithms”
Nichol et al. -.

Large Scale Machine learning

“GraphSAINT Graph Sampling Based Inductive Learning Method”
Zeng et al. - ICLR 2020.

Contribution:
a. Samples sub-graphs and runs GCN on entire sub-graph without node sampling in layers.
b. Defined node and edge sampling procedure in order to avoid bias.

Combinatorial Optimization

“Learning Combinatorial Optimization Algorithms over Graphs”
Dai et al. - NeurIPS 2017.

Contribution:
a. GCN + RL End to End framework for Comb Optimization.
b. Tackled problems like TSP, Vertex Cover, Set Cover etc.

“Combinatorial Optimization with Graph Convolutional Networks and Guided Tree Search”
Lee et al. - NeurIPS 2018.

Contribution:
a. GCN + Guided Tree Search.
b. Parallelizable

“Attention, Learn to Solve Routing Problems!”
Kool et al. - ICLR 2019.

Contribution:
a. GCN + Transformer to encode Graph
b. Decoder to output sequence of nodes.

GNN

“Position-aware Graph Neural Networks”
You et al. - ICML 2019.

Contribution:
a. Two nodes, even if they have same neighboorhood get different embeddings. This is not in the case of GraphSage.
b. Uses Landmark nodes and weighs message from node to landmark nodes using distance to landmark node.

“Hierarchical Graph Representation Learning with Differentiable Pooling”
Ying et al. .

Contribution:
a. Aggregates nodes into clusters.
b. Coarsens clusters into larger level clusters.

“Graph Attention Networks”
Velickovic et al. - ICLR 2018.

Contribution:
a. Aggregate neighbor information based upon importance.
b. Inductive approach/

“Inductive Representation Learning on Large Graphs”
Hamilton et al. - NeurIPS 2017.

Contribution:
a. Inductive approach
b. Learns aggregation matrices.
c. Samples neighorhood uniformly.

Recommendation Systems using Graphs

“Graph Convolutional Neural Networks for Web-Scale Recommender Systems”
Ying et al. - KDD 2018.

“Session-based Social Recommendation via Dynamic Graph Attention Networks”
Song et al. - WSDM 2019.

Contribution:
a. Dynamic user interests and context-dependent social influences.
b. Attention based modelling

“Inductive Matrix Completion Based on Graph Neural Networks”
Zhang et al. - ICLR 2020.

Contribution:
a. Featureless nodes
b. Localized graph- Perform GCN on localized subgraph for user-movie target pair.

Dynamic Networks

“Multi-task Representation Learning for Travel Time Estimation”
Li et al. - KDD 2018.

Contribution:
a. Multi Task framework to predict travel time using auxillary tasks.

“Temporal Network Representation Learning via Historical Neighborhoods Aggregation”
Huang et al. - ICDE 2020.

Contribution:
a. Concept of temporal random walks
b. Different weight to different edges based upon timestamp.
c. Different weightage to different random walks.