CARLA platform enables comprehensive testing of modular, imitation, and reinforcement learning methods for urban autonomous driving, with success rates over 95% in training scenarios.

Alexey Dosovitskiy, German Ros, Felipe Codevilla et al.

Proposes a two-step approach: stochastic dual OT plan learning and neural network Monge map approximation, applied to domain adaptation and generative modeling.

Vivien Seguy, Bharath Bhushan Damodaran, Rémi Flamary et al.

Variational Walkback learns non-equilibrium transition operators to generate samples, excelling on datasets like MNIST.

Anirudh Goyal, Nan Rosemary Ke, Surya Ganguli et al.

Combining trajectory optimization, supervised learning, and model structure to mitigate the curse of dimensionality in bipedal robot control.

Xingye Da, Jessy Grizzle

Proposes SNPE, a neural network-based ABC method for full Bayesian inference of neural models, accurately recovering parameters and handling missing data.

Jan-Matthis Lueckmann, Pedro J. Goncalves, Giacomo Bassetto et al.

Introduces the Fisher-Rao norm to analyze geometry and complexity in neural networks.

Tengyuan Liang, Tomaso Poggio, Alexander Rakhlin et al.

VQ-VAE combines discrete latent variables with vector quantization, achieving near state-of-the-art likelihoods and enabling high-quality multimodal generation.

Aaron van den Oord, Oriol Vinyals, Koray Kavukcuoglu

Hierarchical genetic encoding plus evolution reaches 3.63% CIFAR-10 error and 20.3% top-1 on ImageNet.

Hanxiao Liu, Karen Simonyan, Oriol Vinyals et al.

Grad-CAM++: Incorporates second-order derivatives to improve multi-object localization, achieving 70.72% win rate over Grad-CAM.

Aditya Chattopadhyay, Anirban Sarkar, Prantik Howlader et al.

Constructed boron's interatomic potential using ML and RSS, revealing local energies in β-rhombohedral boron.

Volker L. Deringer, Chris J. Pickard, Gábor Csányi

Proves gradient descent on separable data converges directionally to max-margin (hard margin SVM) solution with rate O(1/ log t).

Daniel Soudry, Elad Hoffer, Mor Shpigel Nacson et al.

Crystal Graph Convolutional Neural Networks (CGCNN) learn material properties directly from atomic connections, achieving high prediction accuracy.

Tian Xie, Jeffrey C. Grossman

Proposes Wasserstein-based distributionally robust regularization, improving model generalization and robustness.

Soroosh Shafieezadeh-Abadeh, Daniel Kuhn, Peyman Mohajerin Esfahani

Proposed a new estimator combining logistic regression and Kullback-Leibler distance to efficiently estimate conversion probability on Criteo dataset.

Abdollah Safari, Rachel MacKay Altman, Thomas M. Loughin

Deep Wasserstein embedding (DWE) learns neural network-based Euclidean approximation of W2 distance, enabling fast large-scale distribution analysis.

Nicolas Courty, Rémi Flamary, Mélanie Ducoffe

Proposes asymmetric Actor-Critic leveraging full states during training and visual inputs at inference, achieving robust sim-to-real transfer without real data.

Lerrel Pinto, Marcin Andrychowicz, Peter Welinder et al.

Constructed WIKIHOP and MEDHOP datasets for multi-hop cross-document QA; models reach 54.5% accuracy, human 85%.

Johannes Welbl, Pontus Stenetorp, Sebastian Riedel

Combining domain randomization with generative autoregressive models, achieving over 90% success on unseen objects in robotic grasping.

Joshua Tobin, Lukas Biewald, Rocky Duan et al.

Proposed LinGapE, a fully adaptive linear bandit pure exploration algorithm with near-optimal sample complexity.

Liyuan Xu, Junya Honda, Masashi Sugiyama

Proposes an incremental deep Boltzmann machine for adaptive hierarchical scene context modeling, dynamically adding layers based on confidence measures.

Fethiye Irmak Doğan, Hande Çelikkanat, Sinan Kalkan