Mario Lanza1
King Abdullah University of Science and Technology1
Mario Lanza1
King Abdullah University of Science and Technology1
Most artificial intelligence systems are based on artificial neural networks (ANNs), which are normally implemented in traditional silicon microchips. These microchips contain complementary metal-oxide-semiconductor (CMOS) circuits that realize different operations, including mathematical operations, digital-to-analogue and analogue-to-digital conversion, and transimpedance amplifiers. One of the most critical operations in ANNs is the vector matrix multiplication (VMM), but realizing it with traditional CMOS circuits consumes too much time and energy. Using crossbar arrays of memristors is much more efficient because the VMM is done in parallel via Ohm’s law and Kirchhoff’s law. In the past few years, multiple studies presented VMM based on memristive devices made of metal-oxides and phase-change materials. Here we present the first VMM operation using two-dimensional (2D) layered materials, and discuss its advantages and challenges compared to traditional memristive materials.