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

With the recent advances in optical phase change material (PCM), photonicin-memory neurocomputing has demonstrated its superiority in optical neuralnetwork (ONN) designs with near-zero static power consumption, time-of-lightlatency, and compact footprint. However, photonic tensor cores require massivehardware reuse to implement large matrix multiplication due to the limitedsingle-core scale. The resultant large number of PCM writes leads to seriousdynamic power and overwhelms the fragile PCM with limited write endurance. Inthis work, we propose a synergistic optimization framework, ELight, to minimizethe overall write efforts for efficient and reliable optical in-memoryneurocomputing. We first propose write-aware training to encourage thesimilarity among weight blocks, and combine it with a post-trainingoptimization method to reduce programming efforts by eliminating redundantwrites. Experiments show that ELight can achieve over 20X reduction in thetotal number of writes and dynamic power with comparable accuracy. With ourELight, photonic in-memory neurocomputing will step forward towards viableapplications in machine learning with preserved accuracy, order-of-magnitudelonger lifetime, and lower programming energy.

ELight: Enabling Efficient Photonic In-Memory Neurocomputing with Life Enhancement