Mots-Clés

Neural networks, behavior, data analysis, Python development

Description

Context : Throughout daily life, we perform complex motor sequences like riding a bike, playing guitar, or speaking, which are gradually acquired through sensorimotor learning via trial and error and repeated practice. This form of learning, known as procedural learning, involves large neural networks in the brain (cortex, basal ganglia, cerebellum), yet the underlying neural mechanisms remain poorly understood. In our team, 'Network Dynamics of Procedural Learning,' we integrate experimental methods (animal conditioning, pharmacological and optogenetic manipulations, neural recordings) with theoretical approaches (behavioral and neural activity analysis, network modeling) to reveal these mechanisms. A key focus is developing Python-based tools to analyze behavioral and neural data. We are seeking a skilled computer scientist to further enhance and adapt these tools to meet the team’s evolving needs. The development will be done in collaboration with team members, both theorists and experimentalists, under the supervision of Julien Braine (PhD in computer science, postdoc in the team for two years), who has led the development of these tools.

Objectives: 
1- Standardizing existing projects to enable cross-project analyses: Currently, each project has its own scripts and analysis pipelines. This setup presents several issues: cross-project analyses are difficult, parts of the code are duplicated, and transferring project ownership can be problematic. A new unified system is being developed. The intern will contribute to improving this system and migrating existing scripts/tools. In doing so, we will apply analyses developed for a specific project to new ones.
2- Developing new analyses and visualizations: We aim to launch more complex analyses, ranging from automatic annotation of abnormal motor behavior using video data to models linking neural activity recorded by hundreds of sensors (>1Gb/min) to behavior (e.g., birdsong, rat movement speed).
3- Collaborating with researchers and students in the team: The intern will interact with various project stakeholders to define needs, establish specifications, and plan the integration of new tools into existing analyses.

Experimental approach: The student will primarily develop in Python, using libraries such as numpy, pandas, and xarray for data computation, matplotlib and seaborn for data visualization, and scipy and sklearn for data processing. Depending on the project, additional specialized libraries like DeepLabCut, SpikeInterface, or DAS may be required. The code will be shared publicly on GitHub and will include thorough documentation.

Scientific and technical skills required for the candidate: The ideal candidate should possess proficiency in Python programming and the numpy and pandas libraries, have a strong foundation in software project management, general knowledge of machine learning (including training, evaluation procedures, metrics, visualizations, and Python tools), experience with matplotlib, seaborn, scipy, sklearn, or xarray, as well as knowledge of databases; familiarity with neurobiology would be an asset but is not required.