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

The increasing interest in TinyML, i.e., near-sensor machine learning onpower budgets of a few tens of mW, is currently pushing toward enablingTinyML-class training as opposed to inference only. Current trainingalgorithms, based on various forms of error and gradient backpropagation, relyon floating-point matrix operations to meet the precision and dynamic rangerequirements. So far, the energy and power cost of these operations has beenconsidered too high for TinyML scenarios. This paper addresses the openchallenge of near-sensor training on a few mW power budget and presents RedMulE- Reduced-Precision Matrix Multiplication Engine, a low-power specializedaccelerator conceived for multi-precision floating-point General Matrix-MatrixOperations (GEMM-Ops) acceleration, supporting FP16, as well as hybrid FP8formats, with {sign, exponent, mantissa}=({1,4,3}, {1,5,2}). We integrateRedMule into a Parallel Ultra-Low-Power (PULP) cluster containing eightenergy-efficient RISC-V cores sharing a tightly-coupled data memory andimplement the resulting system in a 22 nm technology. At its best efficiencypoint (@ 470 MHz, 0.65 V), the RedMulE-augmented PULP cluster achieves 755GFLOPS/W and 920 GFLOPS/W during regular General Matrix-Matrix Multiplication(GEMM), and up to 1.19 TFLOPS/W and 1.67 TFLOPS/W when executing GEMM-Ops,respectively, for FP16 and FP8 input/output tensors. In its best performancepoint (@ 613 MHz, 0.8 V), RedMulE achieves up to 58.5 GFLOPS and 117 GFLOPS forFP16 and FP8, respectively, with 99.4% utilization of the array of ComputingElements and consuming less than 60 mW on average, thus enabling on-devicetraining of deep learning models in TinyML application scenarios whileretaining the flexibility to tackle other classes of common linear algebraproblems efficiently.

RedMule: A Mixed-Precision Matrix-Matrix Operation Engine for Flexible and Energy-Efficient On-Chip Linear Algebra and TinyML Training Acceleration