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 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 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.

Deep RL from human preferences enables complex task learning with less than 1% interaction, using preference-based reward modeling.

Paul Christiano, Jan Leike, Tom B. Brown et al.

Proposes Sinkhorn divergence with entropic smoothing and automatic differentiation for scalable, stable training of generative models, bridging Wasserstein and MMD.

Aude Genevay, Gabriel Peyré, Marco Cuturi

ORGAN combines GAN and reinforcement learning to bias sequence generation towards specific metrics, improving molecular and music sample quality.

Gabriel Lima Guimaraes, Benjamin Sanchez-Lengeling, Carlos Outeiral et al.

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

Nathan Kallus

Parseval networks control spectral norms of layers, improving adversarial robustness while maintaining accuracy, by constraining weights to approximate Parseval tight frames.

Moustapha Cisse, Piotr Bojanowski, Edouard Grave et al.

Virtual Adversarial Training (VAT) uses virtual adversarial loss for regularization in supervised and semi-supervised learning, achieving top performance on SVHN and CIFAR-10.

Takeru Miyato, Shin-ichi Maeda, Masanori Koyama et al.

Proposes importance sampling-based coreset construction, optimizing data reduction for k-means and other ML tasks with theoretical guarantees.

Olivier Bachem, Mario Lucic, Andreas Krause

BOCA algorithm supports continuous multi-fidelity spaces, significantly improving optimization efficiency.

Kirthevasan Kandasamy, Gautam Dasarathy, Jeff Schneider et al.

Introduces R-GCN for knowledge graph completion, achieving 29.8% improvement in link prediction.

Michael Schlichtkrull, Thomas N. Kipf, Peter Bloem et al.

This paper introduces Evolution Strategies (ES) as a scalable alternative to deep RL, achieving near real-time training on thousands of cores with minimal communication.

Tim Salimans, Jonathan Ho, Xi Chen et al.

Proposes generative model-based compressed sensing; if G is L-Lipschitz, O(k log L) Gaussian measurements suffice for near-perfect recovery.

Ashish Bora, Ajil Jalal, Eric Price et al.

Extends Sobolev kernel regularized least squares learning rates using integral operators and embedding properties, improving generalization in hard learning scenarios.

Simon Fischer, Ingo Steinwart

Introduces WGAN using Wasserstein-1 distance to improve training stability and avoid mode collapse in GANs.

Martin Arjovsky, Soumith Chintala, Léon Bottou

Proposes Variational Dropout for deep neural network sparsification, achieving up to 280× parameter reduction with minimal accuracy loss.

Dmitry Molchanov, Arsenii Ashukha, Dmitry Vetrov

Proposes Stein Sample Learning via Stein Variational Gradient Descent (SVGD), enabling neural networks to approximate target distributions without explicit density calculations, applied to deep energy models.

Dilin Wang, Qiang Liu

Proposes a scalable nonparametric Poisson intensity estimation method using transformed RKHS kernels, with theoretical guarantees and efficient computation.

Seth Flaxman, Yee Whye Teh, Dino Sejdinovic

Proposes C2ST, a classifier-based two-sample test with interpretability and strong statistical guarantees.

David Lopez-Paz, Maxime Oquab