English (United Kingdom)

https://curated-unify.zendy.io/wp-json/zendy-region/v1/featured_content/oa?rat=en

https://curated-unify.zendy.io/wp-json/zendy-region/v1/highlighted_journal/

Global: Zendy Plus annual

Presents the access of premium content as premium feature

Premium Content

Presents the keyphrase highlighting as premium feature

Keyphrase Highlighting

Presents the summarisation as premium feature

Summarisation

Insights

Presents the pdf analysis as premium feature

PDF Analysis

Presents the zaia usage as premium feature

ZAIA

Global: Zendy Plus monthly

Global: Zendy Tools annual

Global: Zendy Tools monthly

Global: Zendy Free Plan

Edge computing is a promising solution for handling high-dimensional,multispectral analog data from sensors and IoT devices for applications such asautonomous drones. However, edge devices' limited storage and computingresources make it challenging to perform complex predictive modeling at theedge. Compute-in-memory (CiM) has emerged as a principal paradigm to minimizeenergy for deep learning-based inference at the edge. Nevertheless, integratingstorage and processing complicates memory cells and/or memory peripherals,essentially trading off area efficiency for energy efficiency. This paperproposes a novel solution to improve area efficiency in deep learning inferencetasks. The proposed method employs two key strategies. Firstly, a Frequencydomain learning approach uses binarized Walsh-Hadamard Transforms, reducing thenecessary parameters for DNN (by 87% in MobileNetV2) and enablingcompute-in-SRAM, which better utilizes parallelism during inference. Secondly,a memory-immersed collaborative digitization method is described among CiMarrays to reduce the area overheads of conventional ADCs. This facilitates moreCiM arrays in limited footprint designs, leading to better parallelism andreduced external memory accesses. Different networking configurations areexplored, where Flash, SA, and their hybrid digitization steps can beimplemented using the memory-immersed scheme. The results are demonstratedusing a 65 nm CMOS test chip, exhibiting significant area and energy savingscompared to a 40 nm-node 5-bit SAR ADC and 5-bit Flash ADC. By processinganalog data more efficiently, it is possible to selectively retain valuabledata from sensors and alleviate the challenges posed by the analog data deluge.

Containing Analog Data Deluge at Edge through Frequency-Domain Compression in Collaborative Compute-in-Memory Networks