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 effectiveness of deep neural networks (DNN) in vision, speech, andlanguage processing has prompted a tremendous demand for energy-efficienthigh-performance DNN inference systems. Due to the increasing memory intensityof most DNN workloads, main memory can dominate the system's energy consumptionand stall time. One effective way to reduce the energy consumption and increasethe performance of DNN inference systems is by using approximate memory, whichoperates with reduced supply voltage and reduced access latency parameters thatviolate standard specifications. Using approximate memory reduces reliability,leading to higher bit error rates. Fortunately, neural networks have anintrinsic capacity to tolerate increased bit errors. This can enableenergy-efficient and high-performance neural network inference usingapproximate DRAM devices.  Based on this observation, we propose EDEN, a general framework that reducesDNN energy consumption and DNN evaluation latency by using approximate DRAMdevices, while strictly meeting a user-specified target DNN accuracy. EDENrelies on two key ideas: 1) retraining the DNN for a target approximate DRAMdevice to increase the DNN's error tolerance, and 2) efficient mapping of theerror tolerance of each individual DNN data type to a corresponding approximateDRAM partition in a way that meets the user-specified DNN accuracyrequirements.  We evaluate EDEN on multi-core CPUs, GPUs, and DNN accelerators with errormodels obtained from real approximate DRAM devices. For a target accuracywithin 1% of the original DNN, our results show that EDEN enables 1) an averageDRAM energy reduction of 21%, 37%, 31%, and 32% in CPU, GPU, and two DNNaccelerator architectures, respectively, across a variety of DNNs, and 2) anaverage (maximum) speedup of 8% (17%) and 2.7% (5.5%) in CPU and GPUarchitectures, respectively, when evaluating latency-bound DNNs.

EDEN: Enabling Energy-Efficient, High-Performance Deep Neural Network Inference Using Approximate DRAM