Publications de Marinus Abraham de Vos

Fair Decentralized Learning

Anne-Marie Kermarrec, Rishi Sharma, Marinus Abraham de Vos, Sayan Biswas

Decentralized learning (DL) is an emerging approach that enables nodes to collaboratively train a machine learning model without sharing raw data. In many application domains, such as healthcare, this approach faces challenges due to the high level of hete ...

IEEE2025

Harnessing Increased Client Participation with Cohort-Parallel Federated Learning

Anne-Marie Kermarrec, Rafael Pereira Pires, Akash Balasaheb Dhasade, Marinus Abraham de Vos

Federated learning (FL) is a machine learning approach where nodes collaboratively train a global model. As more nodes participate in a round of FL, the effectiveness of individual model updates by nodes also diminishes. In this study, we increase the effe ...

ACM2025

Boosting Asynchronous Decentralized Learning with Model Fragmentation

Anne-Marie Kermarrec, Rafael Pereira Pires, Rishi Sharma, Marinus Abraham de Vos, Sayan Biswas

Decentralized learning (DL) is an emerging technique that allows nodes on the web to collaboratively train machine learning models without sharing raw data. Dealing with stragglers, i.e., nodes with slower compute or communication than others, is a key cha ...

ACM2025

Practical Federated Learning without a Server

Anne-Marie Kermarrec, Akash Balasaheb Dhasade, Rishi Sharma, Marinus Abraham de Vos, Erick Lavoie

Federated Learning (FL) enables end-user devices to collaboratively train ML models without sharing raw data, thereby preserving data privacy. In FL, a central parameter server coordinates the learning process by iteratively aggregating the trained models ...

ACM2025

Efficient Federated Search for Retrieval-Augmented Generation

Rachid Guerraoui, Anne-Marie Kermarrec, Rafael Pereira Pires, Diana Andreea Petrescu, Marinus Abraham de Vos

Large language models (LLMs) have demonstrated remarkable capabilities across various domains but remain susceptible to hallucinations and inconsistencies, limiting their reliability. Retrieval-augmented generation (RAG) mitigates these issues by grounding ...

ACM2025

Accelerating MoE Model Inference with Expert Sharding

Anne-Marie Kermarrec, Rafael Pereira Pires, Milos Vujasinovic, Marinus Abraham de Vos

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

Fairness Auditing with Multi-Agent Collaboration

Anne-Marie Kermarrec, Akash Balasaheb Dhasade, Marinus Abraham de Vos

Existing work in fairness auditing assumes that each audit is performed independently. In this paper, we consider multiple agents working together, each auditing the same platform for different tasks. Agents have two levers: their collaboration strategy, w ...

IOS Press2024

Energy-Aware Decentralized Learning with Intermittent Model Training

Anne-Marie Kermarrec, Rafael Pereira Pires, Akash Balasaheb Dhasade, Rishi Sharma, Marinus Abraham de Vos

SKIPTRAIN is a novel Decentralized Learning (DL) algorithm, which minimizes energy consumption in decentralized learning by strategically skipping some training rounds and substituting them with synchronization rounds. These trainingsilent periods, besides ...

IEEE2024

LIGHT-HIDRA: Scalable and decentralized resource orchestration in Fog-IoT environments

Marinus Abraham de Vos

With the proliferation of Internet of Things (IoT) ecosystems, traditional resource orchestration mechanisms, executed on fog devices, encounter significant scalability, reliability and security challenges. To tackle these challenges, recent decentralized ...

2024

Noiseless Privacy-Preserving Decentralized Learning

Anne-Marie Kermarrec, Rafael Pereira Pires, Rishi Sharma, Marinus Abraham de Vos, Sayan Biswas

Decentralized learning (DL) enables collaborative learning without a server and without training data leaving the users' devices. However, the models shared in DL can still be used to infer training data. Conventional defenses such as differential privacy ...

Privacy Enhancing Technologies Symposium Advisory Board2024