Pause
Lecture
Moteur de recherche d'offres d'emploi CEA

Analysis of Clients' Rentability in a Distributed Federated Learning System using Blockchain H/F


Vacancy details

General information

Organisation

The French Alternative Energies and Atomic Energy Commission (CEA) is a key player in research, development and innovation in four main areas :
• defence and security,
• nuclear energy (fission and fusion),
• technological research for industry,
• fundamental research in the physical sciences and life sciences.

Drawing on its widely acknowledged expertise, and thanks to its 16000 technicians, engineers, researchers and staff, the CEA actively participates in collaborative projects with a large number of academic and industrial partners.

The CEA is established in ten centers spread throughout France
  

Reference

2021-18452  

Description de l'unité

Within CEA Tech, the technological research department of CEA, the List Institute (CEA List) is dedicated to intelligent digital systems. We have the expertise resulting from a culture of innovation and our mission is to produce and transfer useful technologies to our industrial partners.

The LICIA (Laboratoire systèmes d'Information de Confiance, Intelligents, Auto-organisants) is the CEA List's lab dedicated to study "Trusted distributed systems", and in particular, the blockchain technology.

Position description

Category

Mathematics, information, scientific, software

Contract

Internship

Job title

Analysis of Clients' Rentability in a Distributed Federated Learning System using Blockchain H/F

Subject

The goal of this internship is to analyze the rentability of users in a federated learning setup.

Contract duration (months)

6

Job description

Machine learning is today one of the main fields of study in theoretical computer science. It has many different applications and is useful in our everyday life, for example, in search engines (Google), in translations (Deepl), for feed recommendations (Facebook, Youtube). Some medical services also use machine learning for medical data analysis, controlling inappropriate data, estimating disease breakthroughs, effective monitoring of patients, etc. These services usually collect all the data and use their training algorithms to produce highly specific results.

 

However, in part, due to some privacy reasons, it is becoming more and more interesting to consider Federated Learning (FL), e.g., see Yang et al. (2018), Li et al. (2020). In FL, the data are always on the side of the users. A server is responsible to send to the users a learning model; the users update the model using their data and the learning algorithms, then send back to the server the update. The server then aggregates them to a new model, which is sent back to the users, and the process continues until the end of the learning.

 

The Carnot Project Fanstastyc of CEA List aims to fully decentralize the server role of FL, and therefore to build distributed federated learning applications. To do so, blockchain technology is used to ensure transparency, order, and agreement on the sequence of updates and models.

 

In such systems, the role of all users is essential. If some users decide not to participate, the system may break down. It is therefore important to ensure that the different users have the incentive to participate in the system.

To do so, the successful applicant will analyze the rentability of the different actors, especially, the rentability of the clients.
In more detail, the goal is to (i) show the existence of the different (Nash) equilibria of the game that represents the interactions of the users, (ii) compute these equilibria if they exist, and (iii) if possible, propose incentive mechanisms that induce the existence of a single equilibrium in which the best interest of the users is in line with their participation to the federated learning processes.
The internship is located on the Paris-Saclay campus, at LICIA, the CEA List’s lab dedicated to distributed systems, and blockchain technologies.

Applicant Profile

The candidate must be in “Master 2” or equivalent, in computer science, or related fields.
The candidate is expected to know about theoretical computing and especially (algorithmic) game theory. Knowledge in federated learning and/or in blockchain technologies is appreciated, but not mandatory.

The internship is open to French speakers and to English speakers.

Position location

Site

Saclay

Job location

France

Location

Palaiseau

Candidate criteria

Languages

  • French (Fluent)
  • English (Fluent)

Requester

Position start date

01/03/2022