With the increasing interest in intelligent systems running in heterogeneous environments, federated learning (FL) has emerged as one of the most important paradigms in various machine learning applications. The concept of FL was coined to describe distributed machine learning solutions that collaborate in tackling a main task with the coordination of a central server where data are stored locally and not transferred. Instead, efficient aggregation is achieved by the central server, and this decentralized learning may ensure data privacy for each end-user/device. Hence, each device relies on its own local data for local training, and uploads its model to the centralized server for aggregation prior to send back the aggregated model to all the devices for subsequent local model updates. Existing works tackle FL from different aspects: data partitioning, privacy mechanisms, machine and deep learning models and algorithms, efficient model communication and heterogeneity.

This PhD subject aims at investigating the potential of FL and neural network design for pattern recognition and computer vision tasks. Many subproblems will be considered including FL-driven neural network architecture design, efficient lifelong learning and fusion, FL-driven optimization algorithms, privacy-preserving discriminative and generative neural network architectures, LLMs, etc. Applications will mainly be centered around computer vision tasks including image / video analysis and understanding.

Starting and Duration: the PhD may start as early as January/February 2025 and its duration is 3 years.

Lab: LIP6, CNRS, Sorbonne University, Paris.

Background: we are seeking a highly motivated candidate, with a preferred background in applied mathematics or computer science with more emphasis on statistics, machine / deep learning and/or image processing / pattern recognition, and familiar with existing machine learning tools and programming platforms.

Applications: should be sent to “hichem [dot] sahbi [at] sorbonne-universite [dot] fr” including a CV and all the available university studies transcripts (and if available recommendation letters).