Breaking the von Neumann bottleneck: architecture-level processing-in-memory technology

Zou, Xingqi; Xu, Sheng; Chen, Xiaoming; Ling, Yinghui; Han, Yinhe

doi:10.1007/s11432-020-3227-1

Cited by 97 publications

(49 citation statements)

References 39 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…This investigation is currently the first one employing SNNs for QSAR, proposing this technique as an alternative to classical machine learning or NN methods. Furthermore, exploring neuromorphic computation solutions as alternatives to tackle the von Neumann bottleneck problem [ 57 ] could provide new insights to drive future technologies for drug discovery and virtual screening. In von Neumann’s architecture, the chips move information continuously and at high bandwidth between the central processing unit and memory, wasting time and energy.…”

Section: Discussionmentioning

confidence: 99%

Molecular Toxicity Virtual Screening Applying a Quantized Computational SNN-Based Framework

Nascimben

Rimondini

2023

Molecules

View full text Add to dashboard Cite

Spiking neural networks are biologically inspired machine learning algorithms attracting researchers’ attention for their applicability to alternative energy-efficient hardware other than traditional computers. In the current work, spiking neural networks have been tested in a quantitative structure–activity analysis targeting the toxicity of molecules. Multiple public-domain databases of compounds have been evaluated with spiking neural networks, achieving accuracies compatible with high-quality frameworks presented in the previous literature. The numerical experiments also included an analysis of hyperparameters and tested the spiking neural networks on molecular fingerprints of different lengths. Proposing alternatives to traditional software and hardware for time- and resource-consuming tasks, such as those found in chemoinformatics, may open the door to new research and improvements in the field.

show abstract

Section: Discussionmentioning

confidence: 99%

Molecular Toxicity Virtual Screening Applying a Quantized Computational SNN-Based Framework

Nascimben

Rimondini

2023

Molecules

View full text Add to dashboard Cite

show abstract

“…However, with the rapid development of big data, Internet of Things (IoT), and artificial intelligence (AI), traditional von Neumann computers are facing a series of problems, such as high energy consumption and low processing speed, and they cannot solve complex problems. These challenges are caused by the physical separation of the memory and the processor, which is called the “von Neumann bottleneck” . By contrast, a human brain can process large amounts of information in a parallel manner via its neural networks, which consist of ∼10 11 neurons and ∼10 15 synapses .…”

Section: Introductionmentioning

confidence: 99%

2D-Material-Based Volatile and Nonvolatile Memristive Devices for Neuromorphic Computing

Xia

Huang

Hang

et al. 2023

ACS Materials Lett.

View full text Add to dashboard Cite

Neuromorphic computing can process large amounts of information in parallel and provides a powerful tool to solve the von Neumann bottleneck. Constructing an artificial neural network (ANN) is a common means to realize neuromorphic computing, which has exhibited potential applications in pattern recognition, complex sensing, and other areas. Reservoir computing (RC), which is another approach to realize neuromorphic computing, has shown some progress and attracted researchers' attention. Neuromorphic computing can be generally implemented by fabricating memristive array systems. 2D-material-based memristive systems and their applications in ANN and RC have been investigated substantially in recent years due to the unique properties of these systems, such as atomic-level thickness and high carrier mobility. In this Review, we first discuss the volatility and nonvolatility properties of memristive devices and their applications in ANN and RC. Second, 2D materials that can be used to fabricate these devices are introduced, and their classification, physical properties, and preparation methods are presented. Third, we discuss the working mechanisms of 2D-material-based synaptic devices, the mimicked synaptic functions, and the applications of these devices in neuromorphic computing through ANN and RC. Lastly, the performance, progress, and future development directions of 2D-material-based synaptic devices are analyzed. This work systematically investigates the status of 2D-material-based synaptic devices and promotes their utilization in neuromorphic computing.

show abstract

“…The computing power of the hardware can be solved by increasing the parallelism of the computing unit, while the increase of the computing unit increases the data transmission requirements between the data path and the storage unit. This leads to the problem that the data transmission speed between the two modules is much lower than the calculation speed, that is, the problem of "storage wall" [3,4].…”

Section: Introductionmentioning

confidence: 99%

Memristor-Based Read/Write Circuit with Stable Continuous Read Operation

Bao

Zheng

et al. 2022

Electronics

View full text Add to dashboard Cite

In recent years, computation-intensive applications, such as artificial intelligence, video processing and encryption, have been developing rapidly. On the other hand, the problems of “storage wall” and “power consumption wall” for the traditional storage and computing separated architectures limit the computing performance. The computational circuits and memory cells based on nonvolatile memristors are unified and become a competitive solution to this problem. However, there are various problems that prevent memristor-based circuits from entering practical applications, one of which is the memristor state deviation problem caused by continuous reading. In this paper, we study some circuits studied by predecessors on read/write circuit, compare the experimental results, analyze the reason for the resistance state deviation of memristor, and put forward a new parallel structure of memristor based on opposite polarity. The logic “1” and logic “0” are represented by the positive and negative voltage difference of two memristors with opposite polarity, which can effectively alleviate the problem of the resistance state deviation caused by continuous reading. A reading voltage of 2 V is applied to the four circuits at the same time, and continuous reading is carried out until the output voltage becomes stable. The voltage offset of the optimized circuit when reading logic “0” is reduced to 78 mV, which is significantly smaller than that of other circuits. In addition, when reading logic “1”, it has the effect of enhancing the information stored in the memristor.

show abstract

Breaking the von Neumann bottleneck: architecture-level processing-in-memory technology

Cited by 97 publications

References 39 publications

Molecular Toxicity Virtual Screening Applying a Quantized Computational SNN-Based Framework

Molecular Toxicity Virtual Screening Applying a Quantized Computational SNN-Based Framework

2D-Material-Based Volatile and Nonvolatile Memristive Devices for Neuromorphic Computing

Memristor-Based Read/Write Circuit with Stable Continuous Read Operation

Contact Info

Product

Resources

About