Proposes deep CNN-based point detector MagicPoint and homography estimator MagicWarp for robust, real-time SLAM.

Daniel DeTone, Tomasz Malisiewicz, Andrew Rabinovich

Proposes NASNet architecture via small-scale search on CIFAR-10, transferred to ImageNet, achieving 82.7% top-1 accuracy with 28% fewer FLOPS.

Barret Zoph, Vijay Vasudevan, Jonathon Shlens et al.

Deep Layer Aggregation (DLA) uses iterative and hierarchical fusion to improve recognition with fewer parameters, outperforming traditional skip connections.

Fisher Yu, Dequan Wang, Evan Shelhamer et al.

Large-scale hyperparameter tuning reveals that well-regularized standard LSTM outperforms recent architectures on Penn and Wikitext-2 benchmarks.

Gábor Melis, Chris Dyer, Phil Blunsom

MoCoGAN decomposes content and motion in a latent space, enabling controllable, unsupervised video generation with improved content consistency and dynamic diversity.

Sergey Tulyakov, Ming-Yu Liu, Xiaodong Yang et al.

Proves that over-parameterized shallow neural networks with quadratic activation have no spurious local minima, enabling gradient descent to find global optima efficiently.

Mahdi Soltanolkotabi, Adel Javanmard, Jason D. Lee

Distral enhances multitask reinforcement learning by sharing a distilled policy, excelling in complex 3D environments.

Yee Whye Teh, Victor Bapst, Wojciech Marian Czarnecki et al.

The Deep BSDE method solved 100-dimensional PDEs with 0.17%–0.46% relative error.

Jiequn Han, Arnulf Jentzen, Weinan E

Bayesian, penalization, and model selection methods effectively identify relevant variables in high-dimensional clustering, improving interpretability and accuracy.

Michael Fop, Thomas Brendan Murphy

Proposes a grasp pose detection method in point clouds, achieving a 93% success rate.

Andreas ten Pas, Marcus Gualtieri, Kate Saenko et al.

Introduces the E2E dataset, ten times larger than previous, emphasizing lexical richness, syntactic diversity, and content selection challenges, advancing natural language generation.

Jekaterina Novikova, Ondřej Dušek, Verena Rieser

Introduces continuous-filter convolution (cfconv) layers in SchNet, achieving state-of-the-art energy and force predictions for molecules, overcoming grid limitations.

Kristof T. Schütt, Pieter-Jan Kindermans, Huziel E. Sauceda et al.

Proposes a spectral-norm normalized margin bound for neural networks, showing its correlation with Lipschitz constants and generalization error.

Peter Bartlett, Dylan J. Foster, Matus Telgarsky

Proposes graph-based sampling strategies to accelerate neural network collaborative filtering, achieving up to 30x speedup.

Ting Chen, Yizhou Sun, Yue Shi et al.

Introduces CausE algorithm using causal embeddings to optimize recommendation systems, achieving significant performance improvements.

Stephen Bonner, Flavian Vasile

Proposes a noise-robust Bayesian optimization framework using quasi-Monte Carlo integration, enhancing high-noise, constrained parameter tuning efficiency.

Benjamin Letham, Brian Karrer, Guilherme Ottoni et al.

Proposes a 2D convolution-based point cloud generation framework with differentiable pseudo-rendering for dense 3D reconstruction, outperforming volumetric methods.

Chen-Hsuan Lin, Chen Kong, Simon Lucey

Comparing deep neural networks (DNNs) and humans in object recognition under degraded signals, humans show greater robustness.

Robert Geirhos, David H. J. Janssen, Heiko H. Schütt et al.

Extended first-order method beyond Lipschitz gradient for nonconvex nonsmooth problems, applied to quadratic inverse problems with global convergence guarantees.

Jérôme Bolte, Shoham Sabach, Marc Teboulle et al.

Proposes robust optimization-based adversarial training, achieving over 89% accuracy on MNIST and 46% on CIFAR10 under strong attacks.

Aleksander Madry, Aleksandar Makelov, Ludwig Schmidt et al.