Toward Single‐Cell Multiple‐Strategy Processing Shift Register Powered by Phase‐Change Memory Materials

Abstract

Computers are the foundations of everything from weapons systems to technologies utilized daily by businesses and consumers. The ever-increasing demand for data-intensive applications in business-related fields including machine learning and the internet of things requires highly energy-efficient hardware for operations such as autonomous driving, [1,2] speech recognition, [3,4] image classification, [5] and diseases diagnosis. [6] Moreover, there is an ever-rising demand for readily-accessible-data utilizations in consumer-grade products such as televisions, laptops, and tablets that require advanced technologies, e.g., during the Covid-19 pandemic, as people worked and learned via Zoom, socialized over Skype, and engaged on Netflix to alleviate the lockdown blues. Because these data-intensive/ readily-accessible-data applications require both high performance and energy-efficient operation, the power [7,8] and memory constraints imposed by von Neumann computers, with separate processing and storage units, limit the ability of conventional processors to meet optimal requirement for these applications. [9] Traditional computers utilize a wide range of memory technologies. [10,11] Prototypical register/static random-access memory (SRAM) technologies are fast (few nanoseconds) but volatile and have a low integration density, which limits their use to high-speed/ cache memory applications. [12] The register, based on a set of flip flops, plays a key role in accepting, storing, and transferring data and instructions that are being utilized immediately by the CPU, as well as other digital circuitries. [13] The data can be transferred in a serial manner using shift-based operations. [14] Moreover, conventional dynamic random-access memory (DRAM) technology is moderately fast (several ten nanoseconds) but volatile. [15] As a result, prototypical DRAM requires frequent refresh operations and thus has increased energy consumption. On the other hand, typical flash is nonvolatile but relatively slow (hundreds of microseconds). [11,16]

A Phase‐Change Memristive Reinforcement Learning for Rapidly Outperforming Champion Street‐Fighter Players

“…Finally, the same PCM element can be utilized for both storage and processing, further facilitating hardware cost reduction. [ 66–70 ] Moreover, the previously unexplored training of the SF R2 based on an integrated penalty type was exhibited. The work has also disclosed the utilization of a merged opponent population to train the SF R2, which has not been administered before.…”

“…Finally, the same PCM element can be utilized for both storage and processing, further facilitating hardware cost reduction. [66][67][68][69][70] Moreover, the previously ). The plots disclose that, once the SF R2 obtained the lead in terms of the damage delivered, the SF player could not catch it for most cases.…”

Nonvolatile Memristive Materials and Physical Modeling for In‐Memory and In‐Sensor Computing

“…By expanding to the multilevel-transition realm, in which administering constant stimuli results in a partial, amendable change in the material conductance, a larger degree of freedom was achieved. [86][87][88] For instance, an increased crystallized bit results in an increasing conductance in the PCM layer, wherein the set transition comprises a gradual crystallization of the amorphous volume. As the number of stimuli increases, the PCM layer changes from the amorphous state to the crystallized state.…”

Toward Memristive Phase‐Change Neural Network with High‐Quality Ultra‐Effective Highly‐Self‐Adjustable Online Learning

“…[13][14][15][16][17][18] Computing and data storage have been achieved using memristive hardware through alterable conductance levels, which enables memory and processing to be combined in a parallel-based design. [19][20][21][22][23][24] Experiments have demonstrated the maximum number of distinct conductance levels utilized for neural-network (NN) learning for different memristive elements, e.g., resistive-switching memory (RSM) elements, magnetic-tunnelling memory (MTM) elements, ferroelectric memory (FM) elements, and other memory elements. [25][26][27] A maximum number of distinguished conductance levels used for NN learning of 2-64 has been The T material state for in situ learning.…”