Proposes MMD GAN, integrating adversarial kernel learning to enhance GMMN, achieving superior image quality with smaller batch sizes.

Chun-Liang Li, Wei-Cheng Chang, Yu Cheng et al.

Expert Iteration algorithm combines tree search and deep learning, outperforming REINFORCE in Hex game.

Thomas Anthony, Zheng Tian, David Barber

AVA dataset provides spatio-temporal localization for 80 atomic visual actions, advancing action recognition research.

Chunhui Gu, Chen Sun, David A. Ross et al.

Unified SHAP framework combines six feature attribution methods, ensuring local accuracy, consistency, with improved computational efficiency.

Scott Lundberg, Su-In Lee

Proposes a semiparametric higher-order influence function estimator that avoids density smoothness assumptions by using sample Gram matrix inverses, achieving efficiency under minimal conditions.

Lin Liu, Rajarshi Mukherjee, Whitney K. Newey et al.

Extended multi-armed bandit to non-compliance scenarios, proposing instrument-armed bandits with new regret measures and algorithms.

Nathan Kallus

Introduced Kinetics dataset with 400 action classes, achieving baseline accuracies of up to X%, highlighting bias and diversity issues.

Will Kay, Joao Carreira, Karen Simonyan et al.

Proposed a self-supervised Siamese learning framework for depth estimation in robotic surgery, achieving 0.604 SSI.

Menglong Ye, Edward Johns, Ankur Handa et al.

This paper rigorously proves that under arbitrary interference, unbiased and consistent causal estimators generally do not exist, emphasizing the importance of assumptions about interference structure.

Guillaume Basse, Edoardo Airoldi

Introduces a end-to-end differentiable Key-Value Retrieval Network that outperforms baselines on multi-domain task-oriented dialogue tasks.

Mihail Eric, Christopher D. Manning

Proposed a deep model with intra-attention and reinforcement learning, achieving ROUGE-1 41.16 on CNN/Daily Mail.

Romain Paulus, Caiming Xiong, Richard Socher

Using projected gradient descent, the paper proves linear convergence for learning ReLUs with near-optimal sample complexity in high dimensions.

Mahdi Soltanolkotabi

TriviaQA is a large-scale distant supervision reading comprehension dataset with 650K QA triples, emphasizing complex reasoning and multi-source evidence.

Mandar Joshi, Eunsol Choi, Daniel S. Weld et al.

Proposes three stable deep neural network architectures inspired by ODE stability theory, ensuring robustness in extremely deep models.

Eldad Haber, Lars Ruthotto

Proposes a fully convolutional sequence-to-sequence model with GLU and multi-step attention, outperforming LSTM-based models on WMT tasks with 10x faster training.

Jonas Gehring, Michael Auli, David Grangier et al.

Study compares human and DNN recognition under visual distortions, finding DNNs underperform humans on distorted images.

Samuel Dodge, Lina Karam

InferSent, a supervised framework using SNLI data, achieves superior sentence representation transfer performance.

Alexis Conneau, Douwe Kiela, Holger Schwenk et al.

Constructed ChestX-ray8 dataset with 108,948 images; used weakly-supervised CNNs for multi-label disease detection and localization, achieving 85% accuracy.

Xiaosong Wang, Yifan Peng, Le Lu et al.

Proposes CTRL, a cross-modal temporal regression model for language-based action localization, achieving significant improvements in IoU=0.5 (45.2% R@1) over state-of-the-art.

Jiyang Gao, Chen Sun, Zhenheng Yang et al.

Proposed Dense-Captioning model with multi-scale proposals and contextual captioning, achieving 26.45% B@1 on ActivityNet Captions.

Ranjay Krishna, Kenji Hata, Frederic Ren et al.