Neumann Computer Research Articles

Two‐Dimensional Memtransistors for Non‐Von Neumann Computing: Progress and Challenges

AbstractThe increased demand of high‐performance computing systems has exposed the inherent limitations of the current state‐of‐the‐art von Neumann architecture. Therefore, developing alternate computing primitives that can offer faster computing speed with low energy expenditure is critical. In this context, while several non‐volatile memory (NVM) devices such as synaptic transistors, spintronic devices, phase change memory (PCM), and memristors have been demonstrated in the past, their two‐terminal nature necessitates additional peripheral elements that increase area and energy overhead. Recently, a new multiterminal device prototype known as a memtransistor has shown tremendous potential to overcome these limitations through exceptional control of the gate electrostatics as enabled by 2D channel materials. In this perspective, a brief overview of recent developments in 2D‐memtransistor devices is provided, including their fundamental operational mechanisms and the role of defects in enabling multiple NVM states and optical photoresponse. An overview of their implementation in the context of neuromorphic, probabilistic, information security, and edge‐sensing primitives is also provided. Finally, a futuristic perspective is provided looking toward their successful large‐scale technological integration.

Accurate and efficient molecular dynamics based on machine learning and non von Neumann architecture

Force field-based classical molecular dynamics (CMD) is efficient but its potential energy surface (PES) prediction error can be very large. Density functional theory (DFT)-based ab-initio molecular dynamics (AIMD) is accurate but computational cost limits its applications to small systems. Here, we propose a molecular dynamics (MD) methodology which can simultaneously achieve both AIMD-level high accuracy and CMD-level high efficiency. The high accuracy is achieved by exploiting deep neural network (DNN)’s arbitrarily-high precision to fit PES. The high efficiency is achieved by deploying multiplication-less DNN on a carefully-optimized special-purpose non von Neumann (NvN) computer to mitigate the performance-limiting data shuttling (i.e., ‘memory wall bottleneck’). By testing on different molecules and bulk systems, we show that the proposed MD methodology is generally-applicable to various MD tasks. The proposed MD methodology has been deployed on an in-house computing server based on reconfigurable field programmable gate array (FPGA), which is freely available at http://nvnmd.picp.vip.

An Overview of Neuromorphic Computing for Artificial Intelligence Enabled Hardware-Based Hopfield Neural Network

Compared with von Neumann's computer architecture, neuromorphic systems offer more unique and novel solutions to the artificial intelligence discipline. Inspired by biology, this novel system has implemented the theory of human brain modeling by connecting feigned neurons and synapses to reveal the new neuroscience concepts. Many researchers have vastly invested in neuro-inspired models, algorithms, learning approaches, operation systems for the exploration of the neuromorphic system and have implemented many corresponding applications. Recently, some researchers have demonstrated the capabilities of Hopfield algorithms in some large-scale notable hardware projects and seen significant progression. This paper presents a comprehensive review and focuses extensively on the Hopfield algorithm's model and its potential advancement in new research applications. Towards the end, we conclude with a broad discussion and a viable plan for the latest application prospects to facilitate developers with a better understanding of the aforementioned model in accordance to build their own artificial intelligence projects.

Boolean and Sequential Logic in a One‐Memristor‐One‐Resistor (1M1R) Structure for In‐Memory Computing

AbstractMemristive devices acting as high‐performance in‐memory computing fabrics have attracted much attention for nonvolatile parallel architectures to break the von Neumann bottleneck. Here, for the first time, a nonvolatile digital logic system with three‐terminal Pt/SiO2/Pt/Ag/GeTe/Ta one‐memristor‐one‐resistor (1M1R) structure is presented. Programmable nonvolatile Boolean logic gates and clocked sequential logic blocks including D latch and D flip‐flop are experimentally implemented, based on which a 4‐bit linear‐feedback shift register with timing design is functionally verified in simulation. With information generation, processing, transmission, and storage performing within the same 1M1R‐based logic blocks, these results consolidate the feasibility of building memristive digital computing system for future non‐von Neumann computing.

Atomic Switch: Atom/Ion Movement Controlled Devices for Beyond Von‐Neumann Computers

An atomic switch is a nanoionic device that controls the diffusion of metal ions/atoms and their reduction/oxidation processes in the switching operation to form/annihilate a conductive path. Since metal atoms can provide a highly conductive channel even if their cluster size is in the nanometer scale, atomic switches may enable downscaling to smaller than the 11 nm technology node, which is a great challenge for semiconductor devices. Atomic switches also possess novel characteristics, such as high on/off ratios, very low power consumption and non-volatility. The unique operating mechanisms of these devices have enabled the development of various types of atomic switch, such as gap-type and gapless-type two-terminal atomic switches and three-terminal atomic switches. Novel functions, such as selective volatile/nonvolatile, synaptic, memristive, and photo-assisted operations have been demonstrated. Such atomic switch characteristics can not only improve the performance of present-day electronic systems, but also enable development of new types of electronic systems, such as beyond von- Neumann computers.