Note: When clicking on a Digital Object Identifier (DOI) number, you will be taken to an external site maintained by the publisher.
Some full text articles may not yet be available without a charge during the embargo (administrative interval).
What is a DOI Number?
Some links on this page may take you to non-federal websites. Their policies may differ from this site.
-
Deconvolution is a key component in contemporary neural networks, especially generative adversarial networks (GANs) and fully convolutional networks (FCNs). Due to extra operations of deconvolution compared to convolution, considerable degradation of performance as well as energy efficiency is incurred when implementing deconvolution on the existing resistive random access memory (ReRAM)-based processing-in-memory (PIM) accelerators. In this work, we propose a ReRAM-based accelerator design, RED, for providing high-performance and low-energy deconvolution. We analyze the deconvolution execution on the existing ReRAM-based PIMs and utilize its interior computation pattern for design optimization. RED includes two major contributions: pixel-wise mapping scheme and zero-skipping data flow. Pixel-wise mapping scheme removes the zero insertion and performs convolutions over several ReRAM arrays and thus enables parallel computations with non-zero inputs. Zero-skipping data flow, assisted with customized input buffers design, enhances the computation parallelism and input data reuse. In evaluation, we compare RED against the existing ReRAM-based PIMs and CMOS-based counterpart with a variety of GAN and FCN models, each of which contains multiple deconvolution layers. The experimental results show that RED achieves a 4.0×-56.16× speedup and a 1.05×-18.17× energy efficiency improvement over previous related accelerator designs.more » « less
-
Abstract Whether the stomach influences the progression of nonalcoholic steatohepatitis (NASH) remains largely unknown. Ghrelin, a 28‐amino acid gastric hormone, is critical for the regulation of energy metabolism and inflammation. We investigated whether ghrelin affects the progression of NASH. NASH was induced with lipopolysaccharide (LPS; 240 μg/kg/day) in male C57BL/6J mice with high‐fat diet (HFD). Ghrelin (11 nmol/kg/day) was administrated by a subcutaneous mini‐pump. Liver steatosis, inflammation, and fibrosis were assessed. Kupffer cells and hepatocytes isolated from wild type, GHSR1a−/−or PPARγ+/−mice were cocultured to determine the cellular and molecular mechanism by which ghrelin ameliorates NASH. A low concentration of LPS activates the Kupffer cells, leading to the development of NASH in mice fed HFD. Ghrelin blocked the progression of NASH induced by LPS via GHSR1a‐mediated attenuation of Kupffer cells M1 polarization. GHSR1a was detected in Kupffer cells isolated from wild‐type mice but not in GHSR1a deficient animals. Upon binding with ghrelin, internalization of GHSR1a occurred. Ghrelin reduced levels of tumor necrosis factor‐α and inducible nitricoxide synthase while increasing Arg1 in Kupffer cells treated with LPS. Ghrelin markedly attenuated the upregulation of lipid accumulation induced by the supernatant of Kupffer cells under both basal and LPS‐treated conditions. Deficiency of PPARγ significantly reduced the effect of LPS on the hepatic steatosis in mice and in cultured hepatocytes. Our studies indicate that the stomach may improve the development of NASH via ghrelin. Ghrelin may serve as a marker and therapeutic target for NASH.