Author Image

Hi, I am Clément

Clément Weinreich

Master student at MVA - ENS Paris-Saclay

I am a passionate student at the research master MVA (Mathematics, Vision, Learning) at ENS Paris-Saclay, currently looking for PhD opportunities for 2024. I graduated in 2023 from the Master of Engineering in computer science and cognitive sciences (cognitics) at ENSC - Bordeaux INP. I have research internship experiences in applied mathematics and machine learning at Inria and Ubisoft La Forge, here is an updated CV. Don’t hesitate to check my latest posts where I present some of my research projects. Also, got my first publication Real-Time Neural Materials with Block-Compressed Features accepted at Eurographics 2024!

Skills

Experiences

1
Research Intern
INRIA Research Laboratory

Apr 2024 - Sep 2024, Paris, France

Inria is the French national research institute for digital science and technology. I joined Astra-Vision, the computer vision group of Astra team (in collaboration with Valeo.AI) which is mainly focused on studying 2D vision and 3D perception.

Responsibilities:
  • Developing physics-driven learning approaches to create generalizable and interpretable models for motion dynamics and extreme deformations.

Research Intern
Ubisoft La Forge (research center of Ubisoft)

Feb 2023 - Aug 2023, Bordeaux, France

Ubisoft La Forge is an open research and development initiative that brings together researchers and experts to bridge the gap between academic research and innovation in the video game world.

Responsibilities:
2

3
Applied Mathematics Research Intern
INRIA Research Laboratory

May 2022 - Jul 2022, Bordeaux, France

Inria is the French national research institute for digital science and technology. I joined the project-team ASTRAL which is mainly focused on advanced statistical and probabilistic methods for the analysis and the control of complex stochastic systems.

Responsibilities:
  • Work on dimensionality reduction and variable selection with a thresholded version of the Sliced Inverse Regression (SIR) method.
  • Developed an R package for SIR thresholded (see on github). The package SIRthresholded is published in CRAN.
  • Wrote a vignette explaining the method and how to use the package (see on github pages).
  • Participated to the JDS 2022 conference (largest statistics conference in France). See the slides
  • Co-authoring a paper in the process of being submitted for the Computational Statistics journal.

Robotics Intern
Pollen Robotics

Jun 2021 - Jul 2021, Bordeaux, France

Pollen Robotics is a Startup working on Reachy, an expressive open-source humanoid robot programmable with Python and ROS. Developed with state-of-the-art AI & Robotics technologies, this robot can also be remotely teleoperated with VR.

Responsibilities:
  • Study different methods to improve the grasping operation while teleoperating Reachy.
  • Implement a force regulation algorithm to manage grasping when Reachy is teleoperated with VR.
  • Add haptic feedback to the VR controllers when the user is using Reachy’s grippers.
4

5
Computer Vision Intern
Smartmoov

Apr 2020 - Jul 2020, Rennes, France

Smartmoov is a Startup that enhances learning to drive using an application linked to an in-vehicle camera. The application automatically detects the student’s mistakes and creates statistics on the driving lesson.

Responsibilities:
  • Implement a distance prediction algorithm with a self-supervised monocular depth estimation model (Monodepth2) and an object detection system (YOLOv3).
  • Link the algorithm to the REST API in order to launch the algorithm after a driving lesson.
  • The algorithm is able to monitor compliance with safety distances.

Undergraduate Research Student
IRISA Research Laboratory

Sep 2019 - Apr 2020, Vannes, France

IRISA is a research center of excellence with scientific priorities such as bioinformatics, systems security, new software architectures, virtual reality, big data analysis and artificial intelligence.

Responsibilities:
  • Design and implement a smart multi-hops communication protocol for the LoRa networks in C++.
  • Implement a LoRa network simulator with a TCP socket hub.
6

Education

Master MVA (Mathematics, Vision, Learning)

The MVA (Mathematics, Vision, Learning) is a Master 2 Research program created in 1996 by Robert Azencott. It is run by the Mathematics Department of ENS Paris-Saclay, co-accredited with the Institut Polytechnique de Paris, and shared with the École Nationale des Ponts et Chaussées and Université Paris Cité.

Taken Courses
  • Convex optimization and applications in machine learning
  • Introduction to statistical learning
  • Machine learning for time series
  • Geometric data analysis
  • Introduction to digital imaging
  • Reproducible research
  • Generative models
  • Learning and generation by probability sampling
  • Advanced deep learning
  • 3D Point Clouds and Modelisation
  • Inverse problems and imaging, statistical and stochastic approaches
Exchange Program - 4.00 GPA

UC Davis is a tier-one research university in the US.

Taken Courses
  • Machine Learning
  • Mathematics of Deep Learning
  • Algorithms Design and Analysis
  • Introduction to Aerospace Science Engineering
Master degree of Engineering in Computer Science and Cognitive Sciences

The Graduate School of Cognitive Engineering is a Bordeaux INP engineering school and is a member of the Conférence des Grandes Ecoles. Its unique program in France provides engineers with dual expertise in computer science, IT and in cognitive science. This school puts a strong emphasis on human-centered design, and the importance of keeping the human in the loop.

Taken Courses
  • Applied Mathematics (Statistics, Probability, Optimization, Dynamic Systems)
  • Machine Learning, Deep Learning and symbolic AI
  • Signal and Image Processing
  • Control System
  • Cognitive Engineering
  • User-centered Design
  • Computer Science (Web and Mobile Development, Databases, Programming)
Extracurricular Activities
  • Member of the student arts association
  • Head of the music association
2018-2020
Associate's degree in Computer Science

The University Technology Diploma in Computer science (DUT Informatique) is an intense 2-year training focused on computer science, and applied mathematics for computer science.

Taken Courses
  • Data Structures and Algorithms
  • Advanced programming
  • Discrete Mathematics
  • Calculus
  • Linear Algebra
  • Statistics
  • Graph Theory
  • Introduction to Artificial Intelligence
  • Introduction to Cybersecurity
  • Web and Mobile Development
  • Database programming and administration
  • Networks
  • Operating Systems
Lycée Émile Littré
2015-2018
Higher Secondary School - Baccalaureate in Science with high honours

The baccalauréat is a French national academic qualification obtained at the end of the secondary education. The Science major prepares students to continue their studies in scientific fields such as medicine, engineering and natural sciences.

Projects

Neural Optimal Transport for Generative Modeling
Neural Optimal Transport for Generative Modeling
Co-author March 2024

Theoretical and empirical study of the use of Weak Optimal Transport in generative modeling.

Analysis of 3D Gaussian Splatting
Analysis of 3D Gaussian Splatting
Single author February - March 2024

In depth analysis of the 3DGS method from a conceptual and experimental point of view to understand its limitations and potential improvements.

Latent space oddity, on the curvature of deep generative models
Latent space oddity, on the curvature of deep generative models
Co-author October - December 2023

Conducted a theoretical and experimental analysis of the latent space in Variational Autoencoders using Riemannian geometry.

Texture synthesis based on semi-descrete optimal transport in patch space
Texture synthesis based on semi-descrete optimal transport in patch space
Co-author November 2023 - January 2024

Explored a texture synthesis model that applies local transformations to Gaussian random fields by solving a semi-discrete optimal transport problem on patch space.

Mutlivariate time series clustering based on Common PCA
Mutlivariate time series clustering based on Common PCA
Co-author November 2023 - January 2024

Explored limitations and extensions of a clustering method for multivariate time series based on the Common Principal Component Analysis (CPCA) method.

Deep learning library from scratch
Deep learning library from scratch
Creator February 2022 - April 2022

Neural network library with automatic differentiation made from scratch in Julia.

Colorization with Diffusion models
Colorization with Diffusion models
Group project December 2022

In depth study of the mathematical foundations of conditional diffusion models (see the post on this website). We also started a PyTorch implementation, which is not complete.

R package for variable selection
R package for variable selection
Co-author May - July 2022

R package that allows to perform variable selection with a thresholded version of the Sliced Inverse Regression (SIR) method. (paper in progress)

Augmented reality from scratch
Augmented reality from scratch
Homework November - December 2021

Incrustation of 2D and 3D content on a dynamic surface in a video. To do so, several image processing algorithms have been implemented from scratch in Matlab.

Activation maps extractor tool
Activation maps extractor tool
Creator November 2021

A PyTorch script that can be used to extract the activation maps of a specific, or all the CNN’s layers.

Linear regression with gradient descent from scratch
Linear regression with gradient descent from scratch
Creator February 2021

Implementation of gradient descent algorithms (GD,SGD,BGD,…) for linear regression.

Data analysis project
Data analysis project
Homework June 2021

Study the correlations of the musical and cinematographical tastes of young people.

Artifacts RogueLike
Artifacts RogueLike
Homework April - May 2021

A roguelike game developed in C#. As a space adventurer, you need to find the 3 legendary artifacts as fast as possible!

Quarto in Prolog
Quarto in Prolog
Homework February - March 2022

Implementation of the board game Quarto in Prolog, with 3 levels of AI, including a variant of the MinMax algorithm.

Statistical modeling project
Statistical modeling project
Homework October 2021

Model the sales of a gas station according to the variables available in the dataset.

Survival Issues after Giant Asteroid Impact
Co-author September 2020 - May 2021

Trans-disciplinary research project that has led to a scientific article presented by our tutor Jean-Marc Salotti at the International Astronautical Congress (IAC) 2021 at Dubai.

Some interests

Music

Guitarist for more than 10 years, I co-managed the music association of ENSC from 2020 to 2022. This involved managing a rehearsal room, giving guitar lessons and organizing concerts. I’m always up for jamming with other musicians!

Astronomy

I have always been passionate about space and astrophysical phenomenas. I have a Dobson 254/1200 that makes me able to observe beautiful deep sky objects, but only when it’s not raining in Normandy ;)