Design and modeling of sustainable, autonomous, smart, and low-power connected systems for precision beekeeping

Manager: Dr. Doreid AMMAR

A research project as part of Hugo Hadjur's doctoral thesis, jointly supervised by ENS Lyon and co-directed by Doreid Ammar, Professor and Academic Director of aivancity, and Laurent Lefèvre, Professor and Research Fellow at INRIA/ENS Lyon.
Pollinating insects are threatened with extinction due to the intensity of current agricultural and industrial practices involving pesticides, urban sprawl, pollution, and global warming. This thesis project aims to develop an innovative beekeeping system capable of automatically diagnosing the health of a hive using the Internet of Things and Artificial Intelligence (AI), while ensuring compliance with the constraints of beekeeping and design (maximizing resource use, minimizing energy costs). The development of machine learning algorithms capable of managing multiple data streams generated by a connected hive will be essential to helping protect bees and preserve their role in maintaining ecological balance and biodiversity.
  

ANR SKYDATA Project: Project to study the minimum properties of a distributed system for federated learning

Responsible: Dr. Etienne MAUFFRET

The objective of this research project by the French National Research Agency (ANR), in collaboration with ENS Lyon, is to study and implement an autonomous and intelligent data system. The data in such a system is capable of evolving, migrating, and replicating on its own, without the intervention of a file management system. The use of the latest knowledge in large-scale data management systems, distributed systems, and multi-agent systems with federated learning will facilitate the development of this new data management paradigm.

Fully Distributed Federated Learning in asynchronous dynamic distributed systems

Responsible: Dr. Etienne MAUFFRET, in partnership with: Eddy Caron, Elise Jeanneau, Aurélie Beynier

This project aims to understand and establish effective federated learning algorithms in a fully distributed system. In particular, we will seek to identify the minimum properties required for such systems and determine the performance that can be expected in minimal systems.

Formalization of the Difference-in-Discontinuities methodology

Responsible persons: Dr. Yacine ALLAM and André MIRANDA ACEVEDO

As part of exploratory research, the project aims to develop and formalize a causal inference methodology to address the lack of unbiased and robust methodologies in the field of territorial public policy. The model is constructed in a similar way to a geographic RDD but with a temporal aspect, a sort of mixture of Difference-in-Differences and Discontinuity Regression. More specifically, we seek to isolate the effect of the treatment and estimate it by performing a locally linear regression in the vicinity of a threshold of our explanatory variable.

NLP to identify gender bias in court decisions

Responsible: Dr. Yacine ALLAM

This project aims to use natural language processing (NLP) to analyze potential biases in court decisions. Through the collection of data on court decisions since 1990 (legal databases, archives, websites), the research consists first of extracting key information such as the names of judges, penalties, and final judgments, and then developing methods to process and organize this data for analysis of potential gender bias in sentencing: Do sentences differ depending on the gender of the judge in cases of sexual and gender-based violence? Do sentences differ depending on the defendant's origin?

Diff-in-Disc: A robust method for recovering causality in a spatial context

Responsible: Dr. Yacine ALLAM

This project aims to develop a causal inference methodology, specifically the "diff-in-disc" (difference in discontinuity) method. The main objective is to create a rigorous methodological framework for the diff-in-disc method in a spatial context and to develop a dedicated package.

From sharing to promotion

Assessing the impact of Airbnb's growth on real estate prices by analyzing real estate transaction data from 2018 to 2020

Responsible: Dr. Yacine ALLAM, in partnership with: Julie Le Gallo (Institut Agro Dijon), Marie Breuillé (INRAE)

This research examines the impact of the density of short-term rental (STR) listings on real estate transaction prices in metropolitan France. Using data on Airbnb listings and real estate transaction prices for the period 2018-2020, covering the entire country and thus offering robust external validity, this research focuses on spatial variations within different areas, including major cities, secondary cities, suburbs, and rural areas.

Causal study of discrimination in France by identifying the difference in votes for lists led by candidates of North African origin in the 2014 French municipal elections

Responsible: Dr. Yacine ALLAM
Assessment of the existence and extent of racial discrimination in French municipal elections, by causally analyzing the difference in voting percentages for a list between 2008 and 2014 when there was a change in the head of the list from a non-North African candidate to a North African candidate.

ANR Project: Digital Demonstrators in Higher Education (DemoES)

Responsible: Dr. Doreid AMMAR

aivancity is a partner in the PEIA (Platform for Immersive Learning Experiences) project, led by the Université Polytechnique des Hauts de France, which is one of the 14 universities and three schools selected for the PIA4 DemoES call for proposals. This project aims to tackle new challenges in both teaching and research in higher education and continues the transformation of more inclusive digital teaching practices through collaborative virtual and realistic environments.
The PEIA (Platform for Immersive Learning Experiences) project, led by the Université Polytechnique Hauts-de-France (UPHF) and the Université Catholique de Lille (ICL), proposes the creation of an open-source Immersive Learning Management System (I-LMS). It aims to support an immersive learning community, offer learning experiences in persistent worlds, and publish immersive resources. See: http://peia-demo.com/home Aivancity's involvement in the PEIA project is to participate in the creation of intelligent chatbots as part of the I-LMS. This is a project developed as part of the AI Clinic.

Approximating sparse semi-non-negative matrix factorization for X-ray COVID-19 image classification

Responsible: Dr. Amel MHAMDI

This project was developed in response to the challenges created by the COVID-19 health crisis and its impact on the healthcare system. The main goal is to develop an analysis of reliable digital data on the accurate diagnosis of patients with the disease. The analysis is based on medical imaging data, particularly chest radiology, which was used extensively during the health crisis. The aim of this project is to create a new algorithm, more reliable than current algorithms, enabling the automatic detection of affected patients based on their imaging results. It is within this framework that R&D work has been undertaken.