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

One of the most exciting applications of Spin Torque Magnetoresistive RandomAccess Memory (ST-MRAM) is the in-memory implementation of deep neuralnetworks, which could allow improving the energy efficiency of ArtificialIntelligence by orders of magnitude with regards to its implementation oncomputers and graphics cards. In particular, ST-MRAM could be ideal forimplementing Binarized Neural Networks (BNNs), a type of deep neural networksdiscovered in 2016, which can achieve state-of-the-art performance with ahighly reduced memory footprint with regards to conventional artificialintelligence approaches. The challenge of ST-MRAM, however, is that it is proneto write errors and usually requires the use of error correction. In this work,we show that these bit errors can be tolerated by BNNs to an outstanding level,based on examples of image recognition tasks (MNIST, CIFAR-10 and ImageNet):bit error rates of ST-MRAM up to 0.1% have little impact on recognitionaccuracy. The requirements for ST-MRAM are therefore considerably relaxed forBNNs with regards to traditional applications. By consequence, we show that forBNNs, ST-MRAMs can be programmed with weak (low-energy) programming conditions,without error correcting codes. We show that this result can allow the use oflow energy and low area ST-MRAM cells, and show that the energy savings at thesystem level can reach a factor two.

Implementing Binarized Neural Networks with Magnetoresistive RAM without Error Correction