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

In recent years, the CNNs have achieved great successes in the imageprocessing tasks, e.g., image recognition and object detection. Unfortunately,traditional CNN's classification is found to be easily misled by increasinglycomplex image features due to the usage of pooling operations, hence unable topreserve accurate position and pose information of the objects. To address thischallenge, a novel neural network structure called Capsule Network has beenproposed, which introduces equivariance through capsules to significantlyenhance the learning ability for image segmentation and object detection. Dueto its requirement of performing a high volume of matrix operations, CapsNetshave been generally accelerated on modern GPU platforms that provide highlyoptimized software library for common deep learning tasks. However, based onour performance characterization on modern GPUs, CapsNets exhibit lowefficiency due to the special program and execution features of their routingprocedure, including massive unshareable intermediate variables and intensivesynchronizations, which are very difficult to optimize at software level. Toaddress these challenges, we propose a hybrid computing architecture designnamed \textit{PIM-CapsNet}. It preserves GPU's on-chip computing capability foraccelerating CNN types of layers in CapsNet, while pipelining with an off-chipin-memory acceleration solution that effectively tackles routing procedure'sinefficiency by leveraging the processing-in-memory capability of today's 3Dstacked memory. Using routing procedure's inherent parallellization feature,our design enables hierarchical improvements on CapsNet inference efficiencythrough minimizing data movement and maximizing parallel processing in memory.

Enabling Highly Efficient Capsule Networks Processing Through A PIM-Based Architecture Design