Development of Memristor-based Artificial Synapses for Brain-like Neuromorphic Computer Chips of the Future
Disciplines
Electrical and Computer Engineering
Abstract (300 words maximum)
Neuromorphic computing aims to mimic the brain’s neural architecture to achieve highly efficient, low-power artificial intelligence systems. A key component of this vision is the memristor, a device capable of simulating synaptic plasticity by adjusting its resistance based on past electrical activity. This study investigates whether memristors are viable as artificial synapses to be used in neuromorphic computing. Our research primarily involved learning the theoretical foundations of memristors, understanding their electrical characteristics, and designing simulated circuits using LTSpice. By modeling simple neuromorphic circuits, we analyzed how memristor-based systems can store and process information akin to biological neural networks. While we did not fabricate physical devices, based on our simulations, we can anticipate that the memristors will effectively demonstrate hysteresis loops, which is indicative of artificial synapses and the physical implementation of the memristors is the next goal of the project. This work contributes to the broader discussion of next-generation computing, highlighting the potential of memristors in advancing AI hardware efficiency and brain-inspired learning models. Based on this project's findings, it is believed that using memristors as a vector towards neuromorphic computing has many benefits, such as memory that holds its state even without the presence of electricity, real time adaptation, and improved sensory applications.
Academic department under which the project should be listed
SPCEET - Electrical and Computer Engineering
Primary Investigator (PI) Name
Sandip Das
Development of Memristor-based Artificial Synapses for Brain-like Neuromorphic Computer Chips of the Future
Neuromorphic computing aims to mimic the brain’s neural architecture to achieve highly efficient, low-power artificial intelligence systems. A key component of this vision is the memristor, a device capable of simulating synaptic plasticity by adjusting its resistance based on past electrical activity. This study investigates whether memristors are viable as artificial synapses to be used in neuromorphic computing. Our research primarily involved learning the theoretical foundations of memristors, understanding their electrical characteristics, and designing simulated circuits using LTSpice. By modeling simple neuromorphic circuits, we analyzed how memristor-based systems can store and process information akin to biological neural networks. While we did not fabricate physical devices, based on our simulations, we can anticipate that the memristors will effectively demonstrate hysteresis loops, which is indicative of artificial synapses and the physical implementation of the memristors is the next goal of the project. This work contributes to the broader discussion of next-generation computing, highlighting the potential of memristors in advancing AI hardware efficiency and brain-inspired learning models. Based on this project's findings, it is believed that using memristors as a vector towards neuromorphic computing has many benefits, such as memory that holds its state even without the presence of electricity, real time adaptation, and improved sensory applications.