Description du sujet de thèse

Ring attractor models for navigation implemented in spiking neural networks

Some 10 years ago, in a major development in neuroscience, the head direction neuronal system of the fruit fly was observed. As it was theoretically predicted in the 90 the structure has a ring topology.However, quite unexpectedly, the system encodes direction in a ring of only about 8 neurons!
How such a small system can accurately encode and maintain the information of heading direction of the fly, is a matter of current interest [Noorman et al Nat Neurosci 2024].
In this thesis, the goal will be to understand how the fly accomplishes this cognitive task by implementing a working model in hardware. This work will be based on an innovative methodology that we have been developing in our group in recent years.
We have implemented in analog hardware, circuits of neuron and synapses, which are the basic components to build arbitrary spiking neuronal networks. Our neurons circuits are of extreme simplicity as they are based on the concept of memristors. These are electronic devices that modify their conductance depending on the applied voltage, akin ionic channels in biological neurons.
Our approach is at the crossroads of Physics, Electronics, Neurosciences and Robotics. It is in the framework of our ANR funded project "MemAI", which is a collaboration between Rozenberg group at INCC and Sumbre Lab at IBENS.

The main activities of the student
- design and implement analog circuits that emulate spiking neural networks based on the base circuits already implemented in the group
- verify the design of networks through numerical simulations
- study of the emerging behavior of models through the implementation and measurement of circuits in hardware
- write progress reports and research articles
- make scientific presentations

The desired skills
- experience in electronics, including measurement and PCB design
- numerical simulation
- desire to learn a range of different topics, from neuromorphic features of quantum materials to nonlinear behavior and neuroscience. Prior experience in these areas is desired but not mandatory.

Contexte de travail

The work will be conducted at the INCC of Université Paris Cité in Paris. This is UPC - CNRS lab with research groups in neuroscience, from biology to cognitive psychology. We are a team with an interest in Theoretical Neuroscience, where we aim to understand how neural and cognitive functions emerge from spiking units. Our distinctive feature is that we implement theoretical models in hardware. The research work is in the context of a ANR project in collaboration with the group of G. Sumbre at IBENS