20 March 2025
Neuromorphic computing, inspired by the structure and function of the human brain, aims to emulate biological information processing systems for the development of efficient and brain-like computational frameworks. In a context where a hardware revolution is required to sustain the ever-growing demand for computing power boosted by the development of artificial intelligence, a promising approach relies on exploiting the inherent physics of “intelligent materials” for computation. For this purpose, a wide range of emerging technologies has been demonstrated for in materia computing by emulating hardware neuromorphic functionalities.
In this work, EMERGE partners from the University of Pisa introduce a novel computational framework inspired by the physics of memristive devices and systems, which are embed into the context of Recurrent Neural Networks (RNNs) for time-series processing. The author's proposed memristive-friendly neural network architecture leverages both the principles of Reservoir Computing (RC) and fully trainable RNNs, providing a versatile platform for sequence learning. They provide a mathematical analysis of the stability of the resulting neural network dynamics, identifying the role of crucial RC-based architectural hyper-parameters. Through numerical simulations, they demonstrate the effectiveness of the proposed approach across diverse regression and classification tasks, showcasing performance that is competitive with both traditional RC and fully trainable RNN systems. The results highlight the scalability and adaptability of memristive-inspired computational architectures, offering a promising path toward efficient neuromorphic computing for complex sequence-based applications.
Read the paper in the link below.
More Information
