Low-energy, high-accuracy convolutional network inference in 3D crosspoint (3DXP) arrays

被引：0

作者：

Carletti, F. ^{[1
,2
]}

Farronato, M. ^{[1
,2
]}

Lepri, N. ^{[1
,2
]}

Tortorelli, I ^{[3
]}

Pirovano, A. ^{[3
]}

Fantini, P. ^{[3
]}

Ielmini, D. ^{[1
,2
]}

机构：

[1] Politecn Milan, Dipartimento Elettron Informaz & Bioingn DEIB, Piazza Leonardo da Vinci 32, I-20133 Milan, Italy

[2] IU NET, Piazza Leonardo da Vinci 32, I-20133 Milan, Italy

[3] Micron Technol Inc, Via Trento 26, I-20871 Vimercate, MB, Italy

来源：

2024 50TH IEEE EUROPEAN SOLID-STATE ELECTRONICS RESEARCH CONFERENCE, ESSERC 2024 | 2024年

基金：

欧洲研究理事会;

关键词：

In-memory computing (IMC); 3D crosspoint (3DXP); phase change memory (PCM); artificial intelligence (AI); convolutional neural network (CNN); IMPACT;

D O I：

10.1109/ESSERC62670.2024.10719497

中图分类号：

TM [电工技术]; TN [电子技术、通信技术];

学科分类号：

0808 ; 0809 ;

摘要：

In-memory computing (IMC) has emerged as a promising solution for artificial intelligence (AI) accelerators thanks to the reduced data movement and improved parallelism in the crosspoint memory array. A key issue of IMC is the excessive current of the memory elements causing energy inefficiency and computing inaccuracy due to IR drop. This work reports a hardware demonstration of IMC by a 3D crosspoint (3DXP) array of phase change memory (PCM). We experimentally demonstrate feature extraction, a typical layer of convolutional neural networks (CNNs) and simulate inference of a LeNet CNN for handwritten digits classification (MNIST database). Low energy is enabled by subthreshold operated 3DXP cells, while the high accuracy is supported by precise program-verify algorithms. The impact of read 1/f noise is discussed via measurements and simulations.

引用

页码：412 / 415

页数：4