Mansour Mayaki

Associate Professor (Maître de conférences) — Université Lumière Lyon 2
Member of LIRIS (UMR 5205 CNRS) · Team Imagine
Sustainable / Green AI Efficient deep learning Optimization & compression Time-series anomaly detection

Short bio

Since September 2025, I am an Associate Professor (Maître de conférences) at Université Lumière Lyon 2 and a member of LIRIS. From 2024 to 2025, I was a postdoctoral researcher at Mines Saint-Étienne. I earned my PhD in Computer Science from Université Côte d’Azur and hold engineering degrees from ENSAI Rennes and a Master of Science in Mathematics and Applications Specialization: Pure Mathematics. My research spans deep learning, anomaly detection, computational cost, energy, and sustainable AI.

Research

I develop frugal / Green AI methods to reduce the carbon and energy footprint of machine learning models. I study optimization and compression techniques tailored to resource-constrained settings. In parallel, I work on anomaly and drift detection for time-series data, with applications in health, industry, and the environment.

Topics

  • Efficient deep learning: training/inference cost, energy-aware evaluation, scaling laws
  • Model optimization & compression for constrained deployments
  • Anomaly and drift detection in time series (predictive maintenance, health monitoring)

Collaboration interests

I am particularly interested in collaborations on Green / frugal AI (compression, optimization, deployment), and time-series anomaly detection (health, industry, environment).

Selected publications

A curated list (see full list on Google Scholar and HAL).

  • Modeling Energy Consumption in Deep Learning Architectures Using Power Laws
    M. Mayaki, V. Charpenay — ECAI 2025. · DOI · HAL
  • AnoRand: A Semi-Supervised Deep Learning Anomaly Detection Method by Random Labeling
    M. Mayaki, M. Riveill — 2023 (preprint). · arXiv · HAL
  • Autoregressive based Drift Detection Method
    M. Mayaki, M. Riveill — 2022. · arXiv · IEEE · HAL
  • Machinery Anomaly Detection using artificial neural networks and signature feature extraction
    M. Mayaki, M. Riveill — IJCNN 2023. · IEEE
  • Multiple Inputs Neural Networks for Fraud Detection
    M. Mayaki, M. Riveill — MLCR 2022, pp. 8–13. · DOI · arXiv · HAL
  • Auto-encoder Based Medicare Fraud Detection
    M. Mayaki, M. Riveill — ASPAI 2022. · HAL

See: Google Scholar · HAL

Teaching

  • Introduction to Computer Science & Artificial Intelligence (Université Lumière Lyon 2)
  • Introduction to Computer Science (SKEMA Business School)
  • Management Information Systems (SKEMA Business School)
  • Research project supervision (MSc Data Science & Artificial Intelligence, Université Côte d’Azur)

CV

Download CV (PDF)