CaFPGA: An automatic generation model for CNN accelerator

Convolutional neural networks (CNNs) are gaining considerable popularity in numerous computer-vision applications. A convolutional architecture for fast feature embedding (Caffe) and other general frameworks has been proposed with the development of CNN. The field-programmable gate array (FPGA) as a classical platform is used to accelerate CNNs because CNNs are computationally complex tasks. However, the implementation of CNN on FPGA platforms is difficult. The present study focuses on exploring the performance-resource design space and proposes an automatic generation model to implement the CNN reconfigurable accelerator on the FPGA platform, which uses Caffe description text as its input file. A design-space exploration model is further proposed. This model includes a layer-folding pipeline structure to balance the bandwidth requirements of convolutional and fully connected layers with incremental exploration algorithms to exploit CNN parallelism. The AlexNet, VGG-S, and VGG-16 networks are implemented. The AlexNet accelerator can achieve 593.5 GOPS, and the VGG-16 accelerator can achieve 638.9 GOPS, which is equivalent or even exceeds that of the state-of-the-art CNN accelerator for VGG-16.

Deletion of the epigenetic regulator GcnE in Aspergillus niger FGSC A1279 activates the production of multiple polyketide metabolites

Epigenetic modification is an important regulatory mechanism in the biosynthesis of secondary metabolites in Aspergillus species, which have been considered to be the treasure trove of new bioactive secondary metabolites. In this study, we reported that deletion of the epigenetic regulator gcnE, a histone acetyltransferase in the SAGA/ADA complex, resulted in the production of 12 polyketide secondary metabolites in A. niger FGSC A1279, which was previously not known to produce toxins or secondary metabolites. Chemical workup and structural elucidation by 1D/2D NMR and high resolution electrospray ionization mass (HR-ESIMS) yielded the novel compound nigerpyrone (1) and five known compounds: carbonarone A (2), pestalamide A (3), funalenone (4), aurasperone E (5), and aurasperone A (6). Based on chemical information and the literature, the biosynthetic gene clusters of funalenone (4), aurasperone E (5), and aurasperone A (6) were located on chromosomes of A. niger FGSC A1279. This study found that inactivation of GcnE activated the production of secondary metabolites in A. niger. The biosynthetic pathway for nigerpyrone and its derivatives was identified and characterized via gene knockout and complementation experiments. A biosynthetic model of this group of pyran-based fungal metabolites was proposed.

Comparison of power cycling reliability of flexible PCB interconnect smaller/thinner and larger/thicker power devices with topside Sn-3.5Ag solder joints

The power cycling reliability of flexible printed circuit board (PCB) interconnect smaller/thinner (ST) 9.5 mm × 5.5 mm × 0.07 mm and larger/thicker (LT) 13.5 mm × 13.5 mm × 0.5 mm single Si diode samples have been studied. With the assumption of creep strain accumulation-induced fatigue cracking as the failure mechanism of the Sn-3.5Ag solder joints, finite element (FE) simulations predicted a higher power cycling reliability of soldering the flexible PCB on a ST Si diode than on a LT Si diode under similar power cycling conditions. Then the power cycling test results of 10 samples for each type are reported and discussed. The samples were constructed with commercially available ST Si diodes with 3.2/0.5/0.3 μm thick AlSiCu/NiP/Pd topside metallization and LT Si diodes with 5/0.1/1/1 μm thick Al/Ti/Ni/Ag topside metallization. In contradiction with the FE prediction, most ST Si diode samples were less reliable than those LT Si diode samples. This can be attributed to the fact that the failure of the ST diode samples was associated with the weak bonding and hence the shear-induced local delamination of the topside solder joints from the AlSiCu metallization, while the failure of the LT diode samples was mainly caused by the creep strain accumulation-induced fatigue cracking within the solder joints. Such results can be used to not only provide better understanding of the different failure mechanisms, but also demonstrate the importance of employing an appropriate topside metallization on the power devices.

Re-SET for Transcription

Buffering dosage imbalance of early- and late-replicating genes is important for dividing eukaryotic cells. Voichek et al. (2018) described critical roles of H3K4 methylation and Paf1C in this process, which was regulated by the S phase checkpoint and H3K56 acetylation.

GPR142 prompts glucagon-like Peptide-1 release from islets to improve β cell function

GPR142 agonists are being pursued as novel diabetes therapies by virtue of their insulin secretagogue effects. But it is undetermined whether GPR142's functions in pancreatic islets are limited to regulating insulin secretion. The current study expands research on its action.

H2A.Z Represses Gene Expression by Modulating Promoter Nucleosome Structure and Enhancer Histone Modifications in Arabidopsis

Strengthening die-cast Al-Mg and Al-Mg-Mn alloys with Fe as a beneficial element

Preparation of high strength zirconia by epoxy gel-casting using hydantion epoxy resin as a gelling agent

In this paper, a water soluble hydantion epoxy resin (ER) was employed to prepare zirconia green bodies with high mechanical strength via epoxy gel-casting (EGC) for dental applications. The rheological properties and gelation behaviors of the suspensions, and the microstructure and mechanical strength of zirconia were systematically studied. The results revealed that solid loadings of up to 52.0 vol% could be achieved with the addition of 0.8 wt% dispersant. When the concentration of ER in the 52.0 vol% suspensions increased from 10.0 to 25.0 wt%, the mechanical strength of the resulting green bodies was enhanced from 40.96 ± 5.20 to 51.22 ± 12.01 MPa. A rough sketch of a pre-sintered zirconia crown was soft-machined from the green body consolidated from a 52.0 vol% suspension containing 10.0 wt% ER. The sintered body had a high mechanical strength of 942.77 ± 95.61 MPa.

Ultra-high thermoelectric performance in graphene incorporated Cu2Se: Role of mismatching phonon modes

A thermoelectric material consisting of Cu2Se incorporated with up to 0.45 wt% of graphene nanoplates is reported. The carbon-reinforced Cu2Se exhibits an ultra-high thermoelectric figure-of-merit of zT = 2.44 ± 0.25 at 870 K. Microstructural characterization reveals dense, nanostructured grains of Cu2Se with multilayer-graphene and graphite agglomerations located at grain boundaries. High temperature X-ray diffraction shows that the graphene incorporated Cu2Se matrix retains a cubic structure and the composite microstructure is chemically stable. Based on the experimental structure, density functional theory was used to calculate the formation energy of carbon point defects and the associated phonon density of states. The isolated carbon inclusion is shown to have a high formation energy in Cu2Se whereas graphene and graphite phases are enthalpically stable relative to the solid solution. Neutron spectroscopy proves that there is a frequency mismatch in the phonon density of states between the carbon honeycomb phases and cubic Cu2Se. This provides a mechanism for the strong scattering of phonons at the composite interfaces, which significantly impedes the conduction of heat and enhances thermoelectric performance.

Cubic aggregates of Zn2SnO4 nanoparticles and their application in dye-sensitized solar cells