SIF 3rd Cohort Fellows - Cristian Jimenez Romero, CY Cergy Paris University

• Education

• Experience 

Hybrid Neural Networks: combining Artificial and Spiking Neural networks to enhance life-long continuous learning in autonomous agents and robots

Generalised Continual learning in AI, often referred to as lifelong learning, is the capacity to acquire and adapt to new knowledge and experiences over time, even when dealing with multiple diverse tasks. This area of AI presents a persistent challenge. While there has been notable progress in addressing specific issues, AI systems continue to encounter difficulties in applying what they've learned to unfamiliar tasks. As a result, these systems often excel in specific tasks but face difficulties in multi-task scenarios.

This project aims to enhance continuous learning and adaptability of artificial autonomous systems in dynamic environments. It combines artificial neural networks (ANN) and spiking neural networks (SNN) to create hybrid neural networks (HNN). This integration enables the implementation of learning mechanisms in HNN that closely resemble biological systems.

To this end, we will work along two main hypotheses, namely:
(i) In HNN, subnetworks of SNN are able to combine time-dependent information and increase the adaptability and robustness to changes in the environment, while subnetworks of ANN can provide efficient inference of specific stimuli or patterns.
(ii) The combination of global optimization strategies with biologically inspired plasticity mechanisms in SNN will contribute to the self-organisation of the HNN as a system that allows continuous learning while interacting with a dynamic environment.

The project Led by Dr.Jimenez-Romero and supervised by Prof. Th.Manos (Cergy) with co-supervision by Prof. Y.Timofeeva (Warwick)  integrates current trends in AI, neuroscience, robotics, and complex systems.