Development of neural networks on silicon

An approach called "biomimetic"

where the goal is to imitate the electrical activity of biological neural networks by using models from computational neuroscience. These artificial spiking neurons are connected with biological neurons to study living systems and/or to design solutions for rehabilitation in response to a deficit in a population of neurons in a living organism. So far, these biomimetic systems have been implemented in digital platforms based on FPGA.

An approach called "bio-inspired"

where we use living organisms to create event-based computing systems. This theme uses more formal neural representations. These neural networks are dedicated to RMS tasks (Recognition, Data Mining and Synthesis) for embedded systems. To achieve very low power consumption, these systems combine the bio-inspired design of analog and mixed integrated circuits and nanotechnologies suc ah memristors.

IA teams skills

At the interface of neuroscience and electonic system design, we have developped large skills around our hert :

Embedded real-time computation

Spiking neural networks

Optimization by metaheuristics

Analog and mixed integrated circuit design

Computational neuroscience

Nanotechnologies in neuromorphic systems

Partners

Collaborations and partners

For the various research projects underway, the IMS Bordeaux laboratory and its teams rely on strong partnerships and collaborations, which allow for the creation of a synergy of strengths and a sharing of technical and human resources