The human brain has the ability to perform complex tasks with an energy efficiency far superior to that of the most powerful computers. We took inspiration from the brain to propose neuro-inspired information processing architectures with ultra-low power consumption.
The originality of the IRCICA bio-inspired architecture project is threefold:
• The technology used, the sub-threshold CMOS process, is extremely energy efficient and is able to generate electrical impulses (or spikes) similar in amplitude and time constant to those encountered in biological neural networks.
• The hardware architecture in the form of networks of component networks distributes calculation and storage, which we can be simulated for optimum energy efficiency and speed.
• Natural computation and learning can be simulated and implemented on this type of architecture especially in the field of vision.

The work done in the recent years has identified a number of difficulties to overcome such as: the choice of the pre-processing and input data encoding, the adequacy of learning rules and network architectures to the targeted tasks as well as the up-scaling up (number of layers, quantity and complexity of data).
Future work will be concentrated on three directions: simulation and visualization, unsupervised learning of images and videos, and 3D circuits whose architecture will also be bioinspired.