FEDEREATED LEARNING ON MOBILITY DATA 
POST-DOC - 2 YEARS – TOULOUSE – FRANCE 

Position description

The VILAGIL project focuses on multiple directions to create an ecosystem able of meeting the needs of mobility for the Occitanie region. As part of this project, the "Data and Mobility" action aims to develop mechanisms for automatic integration, multi-store storage and federated access to data. These data will be provided by non-academic project partners such as Toulouse Metropole, Sicoval, Tisséo Collectivité. In some cases this data will be distributed on the different sites of the partners without the possibility of integrating them centrally. The post-doc offer therefore specifically targets this problem by proposing federated learning approaches to allow the exploitation of all this data.
The purpose of this position is to develop aggregation mechanisms for federated learning [1][2]. In FL, organisations can participate in learning tasks without necessarily sharing their data, they just need to share the local settings of the learning models. However, it is not yet clear to provide federated aggregation mechanisms that prove their effectiveness in the presence of heterogeneous data. In the literature, some aggregation methods are proposed such as FedAvg [3], FedMA [4] or FedPer [5]. These methods are limited and do not effectively take into account large amounts of highly heterogeneous data. We would like to be able to propose mechanisms that effectively respond to different aggregation scenarios for different types of learning, given the high heterogeneity of the data that will be collected in VILAGIL (sensor data, images, videos, etc.).

Bibliography
[1] Qiang Yang, Yang Liu, Tianjian Chen, Yongxin Tong. Federated Machine Learning: Concept and Applications. ACM Transactions on Intelligent Systems and Technology 10(2):1-19. 2019. 

[2] Kairouz, P., McMahan, H.B., Avent, B., Bellet, A., Bennis, M., Bhagoji, A.N., Bonawitz, K., Charles, Z., Cormode, G., Cummings, R., D'Oliveira, R.G., Rouayheb, S.E., Evans, D., Gardner, J., Garrett, Z.A., Gascón, A., Ghazi, B., Gibbons, P.B., Gruteser, M., Harchaoui, Z., He, C., He, L., Huo, Z., Hutchinson, B., Hsu, J., Jaggi, M., Javidi, T., Joshi, G., Khodak, M., Konecný, J., Korolova, A., Koushanfar, F., Koyejo, O., Lepoint, T., Liu, Y., Mittal, P., Mohri, M., Nock, R., Özgür, A., Pagh, R., Raykova, M., Qi, H., Ramage, D., Raskar, R., Song, D.X., Song, W., Stich, S.U., Sun, Z., Suresh, A.T., Tramèr, F., Vepakomma, P., Wang, J., Xiong, L., Xu, Z., Yang, Q., Yu, F.X., Yu, H., & Zhao, S. (2019). Advances and Open Problems in Federated Learning. ArXiv, abs/1912.04977.
[3] MCMAHAN, Brendan, MOORE, Eider, RAMAGE, Daniel, et al. Communication-efficient learning of deep networks from decentralized data. In : Artificial Intelligence and Statistics. PMLR, 2017. p. 1273-1282
[4] ARIVAZHAGAN, Manoj Ghuhan, AGGARWAL, Vinay, SINGH, Aaditya Kumar, et al. Federated learning with personalization layers. arXiv preprint arXiv:1912.00818, 2019.
[5] ARIVAZHAGAN, Manoj Ghuhan, AGGARWAL, Vinay, SINGH, Aaditya Kumar, et al. Federated learning with personalization layers. arXiv preprint arXiv:1912.00818, 2019.
Environment
	Laboratory : Institut de Recherche en Information de Toulouse, IRIT, CNRS/UMR5505
	Scientific direction:
Imen Megdiche, SIG Team (Imen.Megdiche@irit.fr)
André Péninou, SIG  Team (Andre.Peninou@irit.fr)
Olivier Teste, SIG Team (Olivier.Teste@irit.fr)
	Location: IRIT Site Paul Sabatier, 118 Route de Narbonne, 31062 Toulouse, FRANCE
	Duration :2 years
	Starting date : March 2021
For Application
We are looking for a motivated candidate with a solid background in applied mathematics and proven knowledge on AI field. Speaking french is an asset. 
 Applications should be sent to  (imen.megdiche@irit.fr) and (olivier.teste@irit.fr) : 
•	CV including scientific publications
•	Cover Letter
•	Copy of the last diploma
•	Report of thesis defence