Publications related to SAMPa: Sharpness-aware Minimization Parallelized

Related publications (20)

Adaptive Test-Time Learning Under Domain Shift: From Augmentation to Prolonged Adaptation

Deep neural networks often fail to generalize when deployed in dynamic and unpredictable real-world environments, where data distributions at test time can differ significantly from those encountered during training. Since simulating all possible distribut ...

EPFL2025

Automated Ensemble Multimodal Machine Learning for Healthcare

Lucas Salomon Brunschwig

The application of machine learning in medicine and healthcare has led to the creation of numerous diagnostic and prognostic models. However, despite their success, current approaches generally issue predictions using data from a single modality. This stan ...

2025

Accelerating MoE Model Inference with Expert Sharding

André Loureiro Espírito Santo, Anne-Marie Kermarrec, Rafael Pereira Pires, Marinus Abraham de Vos, Milos Vujasinovic

Mixture of experts (MoE) models achieve state-of-the-art results in language modeling but suffer from inefficient hardware utilization due to imbalanced token routing and communication overhead. While prior work has focused on optimizing MoE training and d ...

Association for Computing Machinery2025

Label Critic: Design Data Before Models

Andrea Cavalli

As medical datasets rapidly expand, creating detailed annotations of different body structures becomes increasingly expensive and time-consuming. We consider that requesting radiologists to create detailed annotations is unnecessarily burdensome and that p ...

IEEE Computer Society2025

Resource-efficient photonic networks for next-generation AI computing

Demetri Psaltis, Christophe Moser, Niyazi Ulas Dinç, Ilker Oguz, Mustafa Yildirim, Jih-Liang Hsieh

Current trends in artificial intelligence toward larger models demand a rethinking of both hardware and algorithms. Photonics-based systems offer high-speed, energy-efficient computing units, provided algorithms are designed to exploit photonics' unique st ...

SPRINGERNATURE2025

Solving the Cold-Start Problem for the Edge: Clustering and Adaptive Deep Learning for Emotion Detection

José Angel Miranda Calero

Designing AI-based applications personalized to each user's behavior presents significant challenges due to the cold start problem and the impracticality of extensive individual data labeling. These challenges are further compounded when deploying such app ...

IEEE2025

Generative spatial artificial intelligence for sustainable smart cities: A pioneering large flow model for urban digital twin

Paul Erich Keel, Jeffrey Huang, Simon Elias Bibri

Rapid urbanization, alongside escalating resource depletion and ecological degradation, underscores the critical need for innovative urban development solutions. In response, sustainable smart cities are increasingly turning to cutting-edge technologies—su ...

2025

Backpropagation-free Training of Analog AI Accelerators

Romain Christophe Rémy Fleury, Ali Momeni, Matthieu Francis Malléjac

Deep learning has achieved remarkable success in diverse fields in recent years. However, this growth presents significant challenges, particularly in terms of energy consumption during both training and inference phases. While there have been efforts to i ...

Institute of Electrical and Electronics Engineers Inc.2024

In-Silico Trained AI for Enhanced T2Spectrum Imaging and Myelin Water Fraction Mapping in Preclinical 7T MRI

Jean-Philippe Thiran, Erick Jorge Canales Rodriguez, Thomas Yu

This study introduces a Machine Learning (ML) approach for estimating the T2 spectrum and myelin water fraction (MWF) using multi-echo T2 (MET2) data from preclinical 7T Magnetic Resonance Imaging (MRI) scanners. ML methods have shown promise in MWF estima ...

IEEE Computer Society2024

Automatic simulation of electrochemical sensors by machine learning for drugs quantification

Sandro Carrara, Francesca Rodino, Lin Du

Multiple drug concentrations concurrent detection and quantification based on electrochemical sensors are of great importance for therapeutic drug monitoring (TDM) and the development of personalized therapy. Cyclic voltammogram (CV) results obtained by el ...

Pergamon-Elsevier Science Ltd2024

Déjà Vu: Uncanny and Serendipitous Re-encountering of Audiovisual Archives

Yuchen Yang

The advent of the "generative age"marks a pivotal moment in AI evolution, with models like OpenAI's GPT-4 and DALL-E transforming content creation across various fields. These advancements bring unprecedented efficiency and new applications but also presen ...

Association for Computing Machinery, Inc2024

Towards Practical Homomorphic Aggregation in Byzantine-Resilient Distributed Learning

Rachid Guerraoui, John Stephan

The growing availability of distributed data has led to the increased use of machine learning (ML) algorithms in distributed topologies, where multiple nodes collaborate to train models under the coordination of a central server. However, distributed learn ...

Association for Computing Machinery, Inc2024

Multi-Channel Mosra: Mean Opinion Score and Room Acoustics Estimation Using Simulated Data and A Teacher Model

Jozef Marus Coldenhoff

Previous methods for predicting room acoustic parameters and speech quality metrics have focused on the single-channel case, where room acoustics and Mean Opinion Score (MOS) are predicted for a single recording device. However, quality-based device select ...

IEEE2024

Content-Based Objective Evaluation of Artificially Generated Sign Language Videos

Neha Tarigopula

Sign language is vital for communication within the deaf and hard-of-hearing community. Avatar-based methods and deep learning techniques like Generative Adversarial Networks have shown promise in generating sign language video content. One of the challeng ...

IEEE2024

Representation learning algorithms for inferring machine independent latent features in pedestals in JET and AUG

Yoeri Renier Johannes Poels

Variational autoencoder (VAE)-based representation learning algorithms are explored for their capability to disentangle tokamak size dependence from other dependencies in a dataset of thousands of observed pedestal electron density and temperature profiles ...

2024

Optimal Inference in Contextual Stochastic Block Models

Lenka Zdeborová, Odilon Marie Cyprien Duranthon

The contextual stochastic block model (CSBM) was proposed for unsupervised community detection on attributed graphs where both the graph and the high-dimensional node information correlate with node labels. In the context of machine learning on graphs, the ...

Journal of Machine Learning Research Inc.2024

Better Boosting of Communication Oracles, or Not

Artur Riazanov, Nathaniel Idein Harms

Suppose we have a two-party communication protocol for f which allows the parties to make queries to an oracle computing g; for example, they may query an Equality oracle. To translate this protocol into a randomized protocol, we must replace the oracle wi ...

Schloss Dagstuhl- Leibniz-Zentrum fur Informatik GmbH, Dagstuhl Publishing2024

“Flex Tape Can't Fix That”: Bias and Misinformation in Edited Language Models

Anna Sotnikova, Karina Halevy, Antoine Bosselut, Syrielle Mathilde Marie Montariol, Badr Khaled Abdelrahman Alkhamissy

Weight-based model editing methods update the parametric knowledge of language models post-training. However, these methods can unintentionally alter unrelated parametric knowledge representations, potentially increasing the risk of harm. In this work, we ...

Association for Computational Linguistics (ACL)2024

JAX-COSMO: AN END-TO-END DIFFERENTIABLE AND GPU ACCELERATED COSMOLOGY LIBRARY

Austin Chandler Peel

We present jax-cosmo, a library for automatically differentiable cosmological theory calculations. jax-cosmo uses the JAX library, which has created a new coding ecosystem, especially in probabilistic programming. As well as batch acceleration, just-in-tim ...

Maynooth University2023

On the Convergence of Encoder-only Shallow Transformers

Volkan Cevher, Grigorios Chrysos, Fanghui Liu, Yongtao Wu

In this paper, we aim to build the global convergence theory of encoder-only shallow Transformers under a realistic setting from the perspective of architectures, initialization, and scaling under a finite width regime. The difficulty lies in how to tackle ...

2023