VAE combines deep neural networks with probabilistic inference, using reparameterization to efficiently optimize ELBO, enabling high-quality generative modeling.

Diederik P. Kingma, Max Welling

ActiveHARNet integrates Bayesian neural networks and active learning, significantly reducing labeled data points.

Gautham Krishna Gudur, Prahalathan Sundaramoorthy, Venkatesh Umaashankar

Proposes Nyström-based exact DPP sampling with sublinear preprocessing and poly(k) sampling time.

Michał Dereziński, Daniele Calandriello, Michal Valko

Graph Normalizing Flows: a novel reversible GNN model with reduced memory footprint for large-scale graphs.

Jenny Liu, Aviral Kumar, Jimmy Ba et al.

Proves equivalence between graph isomorphism testing and GNN function approximation, introduces Ring-GNN to distinguish non-isomorphic graphs.

Zhengdao Chen, Soledad Villar, Lei Chen et al.

Proposes node- and graph-level pretraining with self-supervised tasks, achieving up to 9.4% ROC-AUC improvement on molecular datasets.

Weihua Hu, Bowen Liu, Joseph Gomes et al.

Proposes a compound scaling method balancing depth, width, and resolution, achieving 84.3% top-1 accuracy on ImageNet with 8.4x fewer parameters.

Mingxing Tan, Quoc V. Le

Cross-lingual knowledge graph alignment using graph matching neural network, significantly improving accuracy.

Kun Xu, Liwei Wang, Mo Yu et al.

Provably Powerful Graph Networks achieves guaranteed 3-WL power with matrix multiplication and reaches 90.55% on MUTAG.

Haggai Maron, Heli Ben-Hamu, Hadar Serviansky et al.

Sherlock employs a multi-input deep neural network trained on 686,765 columns to detect 78 semantic types, achieving an F1 score of 0.89.

Madelon Hulsebos, Kevin Hu, Michiel Bakker et al.

Proves three-layer ResNet can efficiently learn certain functions beyond kernels in a distribution-free setting, with lower sample and computational complexity.

Zeyuan Allen-Zhu, Yuanzhi Li

Introduced CPNGNNs, the first to analyze GNN approximation ratios for combinatorial problems, improving theoretical bounds for dominating set and vertex cover.

Ryoma Sato, Makoto Yamada, Hisashi Kashima

Proposes a batch black-box optimization method with deterministic regret bounds based on frequentist kernel techniques, improving efficiency and robustness.

Yueming Lyu, Yuan Yuan, Ivor W. Tsang

A subexpression-sharing GNN reaches 49.95% proof closure on HOList.

Aditya Paliwal, Sarah Loos, Markus Rabe et al.

MR-GNN enhances structured entity interaction prediction using multi-resolution and dual graph-state LSTM.

Nuo Xu, Pinghui Wang, Long Chen et al.

Cluster-GCN leverages graph clustering to enable efficient deep GCN training on large graphs, reducing memory and computation significantly.

Wei-Lin Chiang, Xuanqing Liu, Si Si et al.

Proposes a DPP-based framework for generating diverse, feasible counterfactual explanations, outperforming prior methods on four datasets.

Ramaravind Kommiya Mothilal, Amit Sharma, Chenhao Tan

MP-DQN fixes P-DQN and wins on all three tasks: 0.987, 0.789, 0.913.

Craig J. Bester, Steven D. James, George D. Konidaris

Proposes weakly-supervised attention for GNNs, achieving over 60% performance improvement on complex graph classification tasks.

Boris Knyazev, Graham W. Taylor, Mohamed R. Amer

Introduces Centered Kernel Alignment (CKA) for measuring neural network representation similarity, overcoming CCA's high-dimensional limitations.

Simon Kornblith, Mohammad Norouzi, Honglak Lee et al.