Towards resilient analog in-memory deep learning via data layout re-organization
Towards resilient analog in-memory deep learning via data layout re-organization
复制标题
通过数据布局重组实现弹性模拟内存深度学习
DOI：
10.1145/3489517.3530532
复制
发表时间：
2022
期刊：
影响因子：
--
通讯作者：
Ewetz, Rickard
中科院分区：
文献类型：
--
作者：
Rashed, Muhammad Rashedul;Awad, Amro;Jha, Sumit Kumar;Ewetz, Rickard
内存处理为神经网络推理机铺平了道路。一个正在出现的挑战是开发软件/硬件接口，以自动将深度学习模型编译到内存计算平台上。深入神经网络(DNN)模型的数据布局组织直接影响模型的分类精度。这是因为交叉开关中的阻性寄生引入了矩阵数据和模拟计算精度之间的依赖关系。为了最大限度地减少寄生的影响，我们首先进行一个案例研究，以了解分别导致低精度和高精度计算的基本矩阵属性。接下来，我们提出了XORG框架，为部署在内存计算平台上的DNN执行数据布局组织。数据布局组织通过在编译时优化交叉开关分配的权重矩阵来提高精度。实验结果表明，XORG框架的准确率提高了3.2倍，平均提高了31%。当使用XORG加速DNN时，写入比特精度要求放宽了1比特，并且提高了对随机电报噪声(RTN)的鲁棒性。
Processing in-memory paves the way for neural network inference engines. An arising challenge is to develop the software/hardware interface to automatically compile deep learning models onto in-memory computing platforms. In this paper, we observe that the data layout organization of a deep neural network (DNN) model directly impacts the model's classification accuracy. This stems from that theresistive parasiticswithin a crossbar introduces a dependency between thematrix dataand theprecisionof the analog computation. To minimize the impact of the parasitics, we first perform a case study to understand the underlying matrix properties that result in computation with low and high precision, respectively. Next, we propose the XORG framework that performs data layout organization for DNNs deployed on in-memory computing platforms. The data layout organization improves precision by optimizing the weight matrix to crossbar assignments at compile time. The experimental results show that the XORG framework improves precision with up to 3.2X and 31% on the average. When accelerating DNNs using XORG, the write bit-accuracy requirements are relaxed with 1-bit and the robustness to random telegraph noise (RTN) is improved.
使用自旋电子学进行可靠的内存中神经形态计算
DOI：
10.1145/3287624.3288745
发表时间：
2019
期刊：
2019 24th Asia and South Pacific Design Automation Conference (ASP-DAC)
影响因子：
--
作者：
Christopher Münch;R. Bishnoi;M. Tahoori
通讯作者：
M. Tahoori