Neural networks in the brain are structured in three-dimensional (3-D) space, and the networks evolve through development and learning, whereas two-dimensional (2-D) crossbars have essentially been optimized for a fully connected neural network, which results in a significant increase in unused memristors. Here, we present a prototype of molecular neural networks on wetware consisting of a space-free synaptic medium immersed in monomer solution. In the medium, conductive polymer wires are grown

more »

... between multiple electrodes through learning only when necessary, i.e., no polymer wire is pre-placed, unlike present 2-D crossbar devices. Through experiments, we found the necessary growth conditions for synaptic polymer wires. We first demonstrated the learning of simple Boolean functions and then data-encoding tasks by using our system comprising the synaptic media and their external controllers. These results are valuable for expanding the concept of space-free synapse development, i.e., extending our 2-D synaptic media to 3-D is possible in principle. Introduction An artificial neural network (ANN) is one of the key components in recent artificial intelligence (AI) based on deep learning technologies, and is capable of performing a variety of computational tasks such as cognition, prediction, optimization, and intuitive representation 1,2 . Present digital computers based on von Neumann architectures are, however, not well-adapted to perform ANN computing because of the physical separation of the arithmetic and memory units 3,4 , which has increased the strong demand for acceleration of AI computing by using special hardware 5 . Among various AI accelerators, including present digital and analog AI accelerators built with silicon complementary metal-oxide semiconductor (CMOS) technologies 6 , a two-dimensional (2-D) array of memristors is one of the promising devices for area-and power-efficient edge-AI