Publications

PETRA: Parallel End-to-end Training with Reversible Architectures

Stephane Rivaud

Louis Fournier

Thomas Pumir

Michael Eickenberg

Edouard Oyallon

Reversible architectures have been shown to be capable of performing on par with their non-reversible architectures, being applied in deep l… (voir plus)earning for memory savings and generative modeling. In this work, we show how reversible architectures can solve challenges in parallelizing deep model training. We introduce PETRA, a novel alternative to backpropagation for parallelizing gradient computations. PETRA facilitates effective model parallelism by enabling stages (i.e., a set of layers) to compute independently on different devices, while only needing to communicate activations and gradients between each other. By decoupling the forward and backward passes and keeping a single updated version of the parameters, the need for weight stashing is also removed. We develop a custom autograd-like training framework for PETRA, and we demonstrate its effectiveness on CIFAR-10, ImageNet32, and ImageNet, achieving competitive accuracies comparable to backpropagation using ResNet-18, ResNet-34, and ResNet-50 models.

2025-01-01

ICLR (publié)

doi.org

openreview.net

Ex Post Conditions for the Exactness of Optimal Power Flow Conic Relaxations

Jean-Luc Lupien

Antoine Lesage-Landry

Convex relaxations of the optimal power flow (OPF) problem provide an efficient alternative to solving the intractable alternating current (… (voir plus)AC) optimal power flow. The conic subset of OPF convex relaxations, in particular, greatly accelerate resolution while leading to high-quality approximations that are exact in several scenarios. However, the sufficient conditions guaranteeing exactness are stringent, e.g., requiring radial topologies. In this short communication, we present two equivalent ex post conditions for the exactness of any conic relaxation of the OPF. These rely on obtaining either a rank-1 voltage matrix or self-coherent cycles. Instead of relying on sufficient conditions a priori, satisfying one of the presented ex post conditions acts as an exactness certificate for the computed solution. The operator can therefore obtain an optimality guarantee when solving a conic relaxation even when a priori exactness requirements are not met. Finally, we present numerical examples from the MATPOWER library where the ex post conditions hold even though the exactness sufficient conditions do not, thereby illustrating the use of the conditions.

2025-01-01

Electric Power Systems Research (publié)

doi.org

arxiv.org

Proceedings of the 18th Workshop on Building and Using Comparable Corpora (BUCC)

Serge Sharoff

Ayla Rigouts Terryn

Pierre Zweigenbaum

Reinhard Rapp

2025-01-01

COLING Workshops (publié)

dblp.uni-trier.de

Pruning Sparse Tensor Neural Networks Enables Deep Learning for 3D Ultrasound Localization Microscopy

Brice Rauby

Paul Xing

Jonathan Porée

Maxime Gasse

Jean Provost

2025-01-01

IEEE Transactions on Image Processing (publié)

doi.org

arxiv.org

RAT: Bridging RNN Efficiency and Attention Accuracy in Language Modeling

Xiuying Wei

Anunay Yadav

Razvan Pascanu

Caglar Gulcehre

Transformers have become the cornerstone of modern large-scale language models; however, their dependence on softmax attention poses a major… (voir plus) computational bottleneck, particularly in long-context settings. In this work, rather than following prevalent approaches such as linear attention (or SSMs) and local attention, we introduce an intermediate design called \rat between recurrence and attention mechanisms. It partitions the input into chunks, applies a simple linear recurrence within each chunk to capture local dependencies, and then performs softmax attention across chunks to model long-range interactions. By adjusting the size of the chunk, \rat enables flexible trade-offs, combining the strengths of RNN and attention. Empirically, with a chunk size of 16, the \rat layer achieves a \(7\times\) improvement in training speed with 100K token sequences and \(9\times\) in generation at 4K sequence length, while maintaining similar or sometimes even better accuracy compared to standard attention. We demonstrate this by training 1.3B parameter models from scratch and performing large-scale evaluations, including short- and long-context benchmarks, as well as supervised fine-tuning~(SFT). We further propose a hybrid architecture that interleaves \rat with local attention. By combining efficient long-range modeling with strong local interactions, this hybrid design not only improves inference speed and reduces cache memory usage compared to attention, but also consistently enhances performance, for example, achieving an average 1 point gain in commonsense reasoning tasks, up to 4 points on code tasks, and a 1 point Rouge-L increase in a summarization SFT task. Code is available at https://github.com/CLAIRE-Labo/RAT

2025-01-01

arXiv.org (prépublication)

doi.org

arxiv.org

Real-time fine finger motion decoding for transradial amputees with surface electromyography

Zihan Weng

Yang Xiao

Peiyang Li

Chanlin Yi

Pouya Bashivan

Hailin Ma

Guang Yao

Yuan Lin

Fali Li

Dezhong Yao 0001

Jingming Hou

Yangsong Zhang

Peng Xu

2025-01-01

Neural Networks (publié)

doi.org

Recovering Dantzig–Wolfe Bounds by Cutting Planes

Ying Chen

Oktay Günlük

Andrea Lodi

Leveraging Dantzig–Wolfe Decomposition in the Original Variable Space for Mixed-Integer Programming Dantzig–Wolfe decomposition has been… (voir plus) extensively applied to solve large-scale mixed-integer programs with decomposable structures, leading to exact solution approaches, such as branch and price. However, these approaches would require solving the problem in an extended variable space and are not readily present in off-the-shelf solvers. In “Recovering Dantzig–Wolfe Bounds by Cutting Planes,” Chen, Günlük, and Lodi propose a computational effective approach for generating cutting planes from Dantzig–Wolfe decomposition to enhance branch and cut in the space of original variables. The proposed approach requires a relatively small number of cutting planes to recover the strength of the Dantzig–Wolfe dual bound and should be easy to implement in general-purpose mixed-integer programming solvers. The authors show that these cutting planes typically lead to a formulation with lower dual degeneracy and hence, a better computational performance than naïve approaches, such as the objective function cut.

2025-01-01

Oper. Res. (publié)

doi.org

arxiv.org

Reinforcement Learning for Versatile, Dynamic, and Robust Bipedal Locomotion Control

Zhongyu Li

Xue Bin Peng

Pieter Abbeel

Sergey Levine

Glen Berseth

Koushil Sreenath

This paper presents a comprehensive study on using deep reinforcement learning (RL) to create dynamic locomotion controllers for bipedal rob… (voir plus)ots. Going beyond focusing on a single locomotion skill, we develop a general control solution that can be used for a range of dynamic bipedal skills, from periodic walking and running to aperiodic jumping and standing. Our RL-based controller incorporates a novel dual-history architecture, utilizing both a long-term and short-term input/output (I/O) history of the robot. This control architecture, when trained through the proposed end-to-end RL approach, consistently outperforms other methods across a diverse range of skills in both simulation and the real world.The study also delves into the adaptivity and robustness introduced by the proposed RL system in developing locomotion controllers. We demonstrate that the proposed architecture can adapt to both time-invariant dynamics shifts and time-variant changes, such as contact events, by effectively using the robot's I/O history. Additionally, we identify task randomization as another key source of robustness, fostering better task generalization and compliance to disturbances. The resulting control policies can be successfully deployed on Cassie, a torque-controlled human-sized bipedal robot. This work pushes the limits of agility for bipedal robots through extensive real-world experiments. We demonstrate a diverse range of locomotion skills, including: robust standing, versatile walking, fast running with a demonstration of a 400-meter dash, and a diverse set of jumping skills, such as standing long jumps and high jumps.

2025-01-01

Int. J. Robotics Res. (publié)

doi.org

arxiv.org

Revisiting Data Augmentation for Ultrasound Images

Adam Tupper

Christian Gagné

Data augmentation is a widely used and effective technique to improve the generalization performance of deep neural networks. Yet, despite o… (voir plus)ften facing limited data availability when working with medical images, it is frequently underutilized. This appears to come from a gap in our collective understanding of the efficacy of different augmentation techniques across different tasks and modalities. One modality where this is especially true is ultrasound imaging. This work addresses this gap by analyzing the effectiveness of different augmentation techniques at improving model performance across a wide range of ultrasound image analysis tasks. To achieve this, we introduce a new standardized benchmark of 14 ultrasound image classification and semantic segmentation tasks from 10 different sources and covering 11 body regions. Our results demonstrate that many of the augmentations commonly used for tasks on natural images are also effective on ultrasound images, even more so than augmentations developed specifically for ultrasound images in some cases. We also show that diverse augmentation using TrivialAugment, which is widely used for natural images, is also effective for ultrasound images. Moreover, our proposed methodology represents a structured approach for assessing various data augmentations that can be applied to other contexts and modalities.

2025-01-01

Trans. Mach. Learn. Res. (publié)

doi.org

openreview.net

Robust prior-biased acquisition function for human-in-the-loop Bayesian optimization

Rose Guay-Hottin

Lison Kardassevitch

Hugo Pham

Guillaume Lajoie

Marco Bonizzato

2025-01-01

Knowl. Based Syst. (publié)