Fully hardware-implemented memristor convolutional neural network.

Memristor-enabled neuromorphic computing systems provide a fast and energy-efficient approach to training neural networks1-4. However, convolutional neural networks (CNNs)-one of the most important models for image recognition5-have not yet been fully hardware-implemented using memristor crossbars, which are cross-point arrays with a memristor device at each intersection. Moreover, achieving software-comparable results is highly challenging owing to the poor yield, large variation and other non-ideal characteristics of devices6-9. Here we report the fabrication of high-yield, high-performance and uniform memristor crossbar arrays for the implementation of CNNs, which integrate eight 2,048-cell memristor arrays to improve parallel-computing efficiency. In addition, we propose an effective hybrid-training method to adapt to device imperfections and improve the overall system performance. We built a five-layer memristor-based CNN to perform MNIST10 image recognition, and achieved a high accuracy of more than 96 per cent. In addition to parallel convolutions using different kernels with shared inputs, replication of multiple identical kernels in memristor arrays was demonstrated for processing different inputs in parallel. The memristor-based CNN neuromorphic system has an energy efficiency more than two orders of magnitude greater than that of state-of-the-art graphics-processing units, and is shown to be scalable to larger networks, such as residual neural networks. Our results are expected to enable a viable memristor-based non-von Neumann hardware solution for deep neural networks and edge computing.

Genome-wide DNA methylation dynamics and RNA-seq analysis during grape (cv. 'Cabernet Franc') skin coloration.

This study generated whole genome DNA methylation maps to characterize DNA methylomes of grape (cv. 'Cabernet Franc') skins and examine their functional significance during grape skin coloration. We sampled grape skin tissues at three key stages (the early stage of grape berry swelling, the late stage of grape berry swelling and the veraison) during which the color of grape berries changed from green to red. DNA methylation levels of grape skins at the three stages were higher in transposable element regions than in the genic regions, and the CG and CHG DNA methylation levels of the genic region were higher than the CHH DNA methylation levels. We identified differentially methylated regions (DMRs) in S2_vs_S1 and S3_vs_S1. The results indicated that DMRs predominantly occurred within the CHH context during grape skin coloration. Many gene ontology (GO)-enriched DMR-related genes were involved in "nucleotide binding," "catalytic activity" and "ribonucleotide binding" terms; however, many KEGG-enriched DMR-related genes were involved in the "flavonoid biosynthesis" pathway. Our results could provide an important foundation for future research on the development mechanism of grape berries.

Overexpressing Ribosomal Protein L16D Affects Leaf Development but Confers Pathogen Resistance in Arabidopsis.

In plant cells, multiple paralogs from ribosomal protein (RP) families are always synchronously expressed, which is likely contributing to ribosome heterogeneity or functional specialization. However, previous studies have shown that most RP mutants share common phenotypes. Consequently, it is difficult to distinguish whether the phenotypes of the mutants have resulted from the loss of specific genes or a global ribosome deficiency. Here, to investigate the role of a specific RP gene, we employed a gene overexpression strategy. We found that Arabidopsis lines overexpressing RPL16D (L16D-OEs) display short and curled rosette leaves. Microscopic observations reveal that both the cell size and cell arrangement are affected in L16D-OEs. The severity of the defect is positively correlated with RPL16D dosage. By combining transcriptomic and proteomic profiling, we found that overexpressing RPL16D decreases the expression of genes involved in plant growth, but increases the expression of genes involved in immune response. Overall, our results suggest that RPL16D is involved in the balance between plant growth and immune response.

Design, analysis and optimization of a waveguide-type near-eye display using a pin-mirror array and a concaved reflector.

Waveguides have become one of the most promising optical combiners for see-through near-eye displays due to the thickness, weight, and transmittance. In this study, we propose a waveguide-type near-eye display using a pin-mirror array and a concaved reflector with a compact outlook, optimized image uniformity and stray light. Issues have been discussed in detail, which include field of view (FOV), eye-box, resolution, depth of field (DOF), display uniformity and stray light artifacts. It can be shown that the DOF can be extended (when compared with traditional waveguide-type near-eye displays) to alleviate the vergence-accommodation conflict (VAC) problem, and the uniformity & stray light can be improved with an optimal structure. Moreover, reflective surfaces have been introduced as the input and output coupling with a compact outlook, an easy-processing structure and the achromatic performance. A prototype based on the proposed method have been successfully developed, and virtual images with an extended DOF can be shown along with the real-world.

Design of a near-eye display measurement system using an anthropomorphic vision imaging method.

We developed a new near-eye display measurement system using anthropomorphic vision imaging to measure the key parameters of near-eye displays, including field-of-view (FOV), angular resolution, eye box, and virtual image depth. The characteristics of the human eye, such as pupil position, pupil size variation, accommodation function, and the high resolution of the fovea, are imitated by the proposed measurement system. A FOV scanning structure, together with a non-vignetting image-telecentric lens system, captures the virtual image from the near-eye display by imitating human eye function. As a proof-of-concept, a prototype device was used to obtain large-range, high-resolution measurements for key parameters of near-eye displays.

A hierarchical sparse coding model predicts acoustic feature encoding in both auditory midbrain and cortex.

The auditory pathway consists of multiple stages, from the cochlear nucleus to the auditory cortex. Neurons acting at different stages have different functions and exhibit different response properties. It is unclear whether these stages share a common encoding mechanism. We trained an unsupervised deep learning model consisting of alternating sparse coding and max pooling layers on cochleogram-filtered human speech. Evaluation of the response properties revealed that computing units in lower layers exhibited spectro-temporal receptive fields (STRFs) similar to those of inferior colliculus neurons measured in physiological experiments, including properties such as sound onset and termination, checkerboard pattern, and spectral motion. Units in upper layers tended to be tuned to phonetic features such as plosivity and nasality, resembling the results of field recording in human auditory cortex. Variation of the sparseness level of the units in each higher layer revealed a positive correlation between the sparseness level and the strength of phonetic feature encoding. The activities of the units in the top layer, but not other layers, correlated with the dynamics of the first two formants (F1, F2) of all phonemes, indicating the encoding of phoneme dynamics in these units. These results suggest that the principles of sparse coding and max pooling may be universal in the human auditory pathway.

Utilization of species checklist data in revealing the spatial distribution of fish diversity.

Marine fish species checklists from six Chinese coastal waters were combined for the analysis of taxonomic diversity. The Genus-Family index (G-F index) ranged between 0.39 and 0.84, which generally indicated a decreasing trend with increasing latitude, with the exception of the southernmost area. Average taxonomic distinctness showed a slight increasing trend from northern to central study areas, but whether the taxonomic distinctness indices represent a latitudinal gradient of biodiversity requires further study. The multivariate analysis revealed a distinct latitudinal variation in fish assemblages. These results indicate that species checklist data are helpful in understanding the diversity distribution of fish species in the coastal zone. The potential of a species inventory should be exploited to fully understand biodiversity.

DRL1, Encoding A NAC Transcription Factor, Is Involved in Leaf Senescence in Grapevine.

The NAC (for NAM, ATAF1,2, and CUC2) proteins family are plant-specific transcription factors, which play important roles in leaf development and response to environmental stresses. In this study, an NAC gene, DRL1, isolated from grapevine Vitis vinifera L. "Yatomi Rose", was shown to be involved in leaf senescence. The quantity of DRL1 transcripts decreased with advancing leaf senescence in grapevine. Overexpressing the DRL1 gene in tobacco plants significantly delayed leaf senescence with respect to chlorophyll concentration, potential quantum efficiency of photosystem II (Fv/Fm), and ion leakage. Moreover, exogenous abscisic acid (ABA) markedly reduced the expression of DRL1, and the ABA and salicylic acid (SA) concentration was lower in the DRL1-overexpressing transgenic plants than in the wild-type plants. The DRL1 transgenic plants exhibited reduced sensitivity to ABA-induced senescence but no significant change in the sensitivity to jasmonic acid-, SA- or ethylene-induced senescence. Transcriptomic analysis and RNA expression studies also indicated that the transcript abundance of genes associated with ABA biosynthesis and regulation, including 9-cis-epoxycarotenoid dioxygenase (NCED1), NCED5, zeaxanthin epoxidase1 (ZEP1), ABA DEFICIENT2 (ABA2), ABA4, and ABA INSENSITIVE 2 (ABI2), was markedly reduced in the DRL1-overexpressing plants. These results suggested that DRL1 plays a role as a negative regulator of leaf senescence by regulating ABA synthesis.

Structural analysis of the essential environmental factors associated with the marine meiofauna.

A biogeochemical study was carried out from 2006 to 2007 in Bohai Bay, northern China, to investigate meiofauna and a number of environmental factors. A total of 18 environmental factors were selected to study their interrelations using Interpretive Structural Analysis. The results showed that these environmental factors could be divided into five levels according to their degree of impact; the results also showed clear links among these factors. The six factors designated as Level 1 included both biological and chemical factors. The other factors had indirect effects on meiofauna and affect these six factors in different ways. There was single factor, sediment grain size, in the fifth level. The structural analysis of the essential factors gave us more information than simple correlation analysis did. Both the water and sediment factors influenced the benthos community. The result is helpful in understanding the complex relationships between environmental factors, and could be used to grasp the main causes and their environmental linkages with further studies; it would be a good basis for integrative environmental management.

Adjusting the structure and luminescence properties of Sr2-x Bax MgAl22 O36 :Eu2+ phosphors by Sr:Ba ratio.

Europium ion (Eu2+ )-doped phosphors exhibit adjustable photoluminescence due to the sensitivity of their luminescence to the local environment. It is of great significance to adjust the luminescence of Eu2+ by changing their local environment in the host. In this work, we investigated the effect of strontium/barium (Sr:Ba) ratio on the structure and luminescence properties of Sr2-x Bax MgAl22 O36 :Eu2+ phosphors. Our investigation indicates that with the decrease of Sr:Ba ratio, the matrix lattice gradually expands and the peak wavelength for the luminescence of Eu2+ presents an obvious blue shift. The occupancy of Eu2+ was analyzed and the reason for the blue shift was explained. Thermal stability for the luminescence of Eu2+ can also be adjusted by changing the Sr:Ba ratio. This work has a positive effect on the regulation of the emission of phosphors and the improvement of thermal stability, which will promote the application of Sr2-x Bax MgAl22 O36 :Eu2+ phosphors in the field of white light emitting diodes.

A R2R3-MYB Transcription Factor, VvMYBC2L2, Functions as a Transcriptional Repressor of Anthocyanin Biosynthesis in Grapevine (Vitis vinifera L.).

In grapevine, the MYB transcription factors play an important role in the flavonoid pathway. Here, a R2R3-MYB transcription factor, VvMYBC2L2, isolated from Vitis vinifera cultivar Yatomi Rose, may be involved in anthocyanin biosynthesis as a transcriptional repressor. VvMYBC2L2 was shown to be a nuclear protein. The gene was shown to be strongly expressed in root, flower and seed tissue, but weakly expressed during the fruit development in grapevine. Overexpressing the VvMYBC2L2 gene in tobacco resulted in a very marked decrease in petal anthocyanin concentration. Expression analysis of flavonoid biosynthesis structural genes revealed that chalcone synthase (CHS), dihydroflavonol 4-reductase (DFR), leucoanthocyanidin reductase (LAR) and UDP glucose flavonoid 3-O-glucosyl transferase (UFGT) were strongly down-regulated in the VvMYBC2L2-overexpressed tobacco. In addition, transcription of the regulatory genes AN1a and AN1b was completely suppressed in transgenic plants. These results suggested that VvMYBC2L2 plays a role as a negative regulator of anthocyanin biosynthesis.

Spin-polarized transport in graphene nanoribbons with Rashba spin-orbit interaction: the effects of spatial symmetry.

We study the spin-dependent transport properties of graphene nanoribbons with Rashba spin-orbit interaction (SOI). It is found that highly spin polarized electrical currents can be produced in asymmetrically-notched graphene nanoribbons, and the polarization components are found to be along the x, y and z directions. The spin polarization is largely enhanced by breaking the spatial symmetries of ideal graphene nanoribbons with Rashba SOI, and the spin polarized electrical currents with higher flexibility in the orientation of the polarization can be generated. This offers new possibilities for the generation of high spin polarization in graphene nanoribbons without external magnetic fields.

Deciphering phonemes from syllables in blood oxygenation level-dependent signals in human superior temporal gyrus.

Linguistic units such as phonemes and syllables are important for speech perception. How the brain encodes these units is not well understood. Many neuroimaging studies have found distinct representations of consonant-vowel syllables that shared one phoneme and differed in the other phoneme (e.g. /ba/ and /da/), but it is unclear whether this discrimination ability is due to the neural coding of phonemes or syllables. We combined functional magnetic resonance imaging with multivariate pattern analysis to explore this question. Subjects listened to nine Mandarin syllables in a consonant-vowel form. We successfully decoded phonemes from the syllables based on the blood oxygenation level-dependent signals in the superior temporal gyrus (STG). Specifically, a classifier trained on the cortical patterns elicited by a set of syllables, which contained two phonemes, could distinguish the cortical patterns elicited by other syllables that contained the two phonemes. The results indicated that phonemes have unique representations in the STG. In addition, there was a categorical effect, i.e. the cortical patterns of consonants were similar, and so were the cortical patterns of vowels. Further analysis showed that phonemes exhibited stronger encoding specificity in the mid-STG than in the anterior STG.

Klein Tunneling in ß12 Borophene.

Motivated by the recent observation of Klein tunneling in 8-Pmmn borophene, we delve into the phenomenon in ß12 borophene by employing tight-binding approximation theory to establish a theoretical mode. The tight-binding model is a semi-empirical method for establishing the Hamiltonian based on atomic orbitals. A single cell of ß12 borophene contains five atoms and multiple central bonds, so it creates the complexity of the tight-binding model Hamiltonian of ß12 borophene. We investigate transmission across one potential barrier and two potential barriers by changing the width and height of barriers and the distance between two potential barriers. Regardless of the change in the barrier heights and widths, we find the interface to be perfectly transparent for normal incidence. For other angles of incidence, perfect transmission at certain angles can also be observed. Furthermore, perfect and all-angle transmission across a potential barrier takes place when the incident energy approaches the Dirac point. This is analogous to the "super", all-angle transmission reported for the dice lattice for Klein tunneling across a potential barrier. These findings highlight the significance of our theoretical model in understanding the complex dynamics of Klein tunneling in borophene structures.

De novo assembling a high-quality genome sequence of Amur grape (Vitis amurensis Rupr.) gives insight into Vitis divergence and sex determination.

To date, there has been no high-quality sequence for genomes of the East Asian grape species, hindering biological and breeding efforts to improve grape cultivars. This study presents ~522 Mb of the Vitis amurensis (Va) genome sequence containing 27 635 coding genes. Phylogenetic analysis indicated that Vitis riparia (Vr) may have first split from the other two species, Va and Vitis vinifera (Vv). Divergent numbers of duplicated genes reserved among grapes suggests that the core eudicot-common hexaploidy (ECH) and the subsequent genome instability still play a non-negligible role in species divergence and biological innovation. Prominent accumulation of sequence variants might have improved cold resistance in Va, resulting in a more robust network of regulatory cold resistance genes, explaining why it is extremely cold-tolerant compared with Vv and Vr. In contrast, Va has preserved many fewer nucleotide binding site (NBS) disease resistance genes than the other grapes. Notably, multi-omics analysis identified one trans-cinnamate 4-monooxygenase gene positively correlated to the resveratrol accumulated during Va berry development. A selective sweep analysis revealed a hypothetical Va sex-determination region (SDR). Besides, a PPR-containing protein-coding gene in the hypothetical SDR may be related to sex determination in Va. The content and arrangement order of genes in the putative SDR of female Va were similar to those of female Vv. However, the putative SDR of female Va has lost one flavin-containing monooxygenase (FMO) gene and contains one extra protein-coding gene uncharacterized so far. These findings will improve the understanding of Vitis biology and contribute to the improvement of grape breeding.

Arabidopsis NOTCHLESS plays an important role in root and embryo development.

Ribosome biogenesis is a fundamental process in eukaryotic cells. NOTCHLESS (NLE) is involved in 60S ribosome biogenesis in yeast, but its role in Arabidopsis (A. thaliana) remains exclusive. Here, we found that Arabidopsis NLE (AtNLE) is highly conservative in phylogeny, which encoding a WD40-repeat protein. AtNLE is expressed in actively dividing tissues. AtNLE-GFP is localized in the nucleus. AtNLE physically interacts with the MIDAS domain of AtMDN1, a protein involved in the biogenesis of the 60S ribosomal subunit in Arabidopsis. The underexpressing mutant nle-2 shows short roots and reduced cell number in the root meristem. In addition, the null mutant nle-1 is embryo lethal, and defective embryos are arrested at the early globular stage. This work suggests that AtNLE interacts with AtMDN1, and AtNLE functions in root and embryo development.

Monolithic 3D Integration of Analog RRAM-Based Computing-in-Memory and Sensor for Energy-Efficient Near-Sensor Computing.

In the era of the Internet of Things, vast amounts of data generated at sensory nodes impose critical challenges on the data-transfer bandwidth and energy efficiency of computing hardware. A near-sensor computing (NSC) architecture places the processing units closer to the sensors such that the generated data can be processed almost in situ with high efficiency. This study demonstrates the monolithic three-dimensional (M3D) integration of a photosensor array, analog computing-in-memory (CIM), and Si complementary metal-oxide-semiconductor (CMOS) logic circuits, named M3D-SAIL. This approach exploits the high-bandwidth on-chip data transfer and massively parallel CIM cores to realize an energy-efficient NSC architecture. The 1st layer of the Si CMOS circuits serves as the control logic and peripheral circuits. The 2nd layer comprises a 1 k-bit one-transistor-one-resistor (1T1R) array with InGaZnOx field-effect transistor (IGZO-FET) and resistive random-access memory (RRAM) for analog CIM. The 3rd layer comprises multiple IGZO-FET-based photosensor arrays for wavelength-dependent optical sensing. The structural integrity and function of each layer are comprehensively verified. Furthermore, NSC is implemented using the M3D-SAIL architecture for a typical video keyframe-extraction task, achieving a high classification accuracy of 96.7% as well as a 31.5× lower energy consumption and 1.91× faster computing speed compared to its 2D counterpart.

Edge learning using a fully integrated neuro-inspired memristor chip.

Learning is highly important for edge intelligence devices to adapt to different application scenes and owners. Current technologies for training neural networks require moving massive amounts of data between computing and memory units, which hinders the implementation of learning on edge devices. We developed a fully integrated memristor chip with the improvement learning ability and low energy cost. The schemes in the STELLAR architecture, including its learning algorithm, hardware realization, and parallel conductance tuning scheme, are general approaches that facilitate on-chip learning by using a memristor crossbar array, regardless of the type of memristor device. Tasks executed in this study included motion control, image classification, and speech recognition.

Electro-mechano responsive elastomers with self-tunable conductivity and stiffness.

Materials with programmable conductivity and stiffness offer new design opportunities for next-generation engineered systems in soft robotics and electronic devices. However, existing approaches fail to harness variable electrical and mechanical properties synergistically and lack the ability to self-respond to environmental changes. We report an electro-mechano responsive Field's metal hybrid elastomer exhibiting variable and tunable conductivity, strain sensitivity, and stiffness. By synergistically harnessing these properties, we demonstrate two applications with over an order of magnitude performance improvement compared to state-of-the-art, including a self-triggered multiaxis compliance compensator for robotic manipulators, and a resettable, highly compact, and fast current-limiting fuse with an adjustable fusing current. We envisage that the extraordinary electromechanical properties of our hybrid elastomer will bring substantial advancements in resilient robotic systems, intelligent instruments, and flexible electronics.

Monolithic three-dimensional integration of RRAM-based hybrid memory architecture for one-shot learning.

In this work, we report the monolithic three-dimensional integration (M3D) of hybrid memory architecture based on resistive random-access memory (RRAM), named M3D-LIME. The chip featured three key functional layers: the first was Si complementary metal-oxide-semiconductor (CMOS) for control logic; the second was computing-in-memory (CIM) layer with HfAlOx-based analog RRAM array to implement neural networks for feature extractions; the third was on-chip buffer and ternary content-addressable memory (TCAM) array for template storing and matching, based on Ta2O5-based binary RRAM and carbon nanotube field-effect transistor (CNTFET). Extensive structural analysis along with array-level electrical measurements and functional demonstrations on the CIM and TCAM arrays was performed. The M3D-LIME chip was further used to implement one-shot learning, where ~96% accuracy was achieved on the Omniglot dataset while exhibiting 18.3× higher energy efficiency than graphics processing unit (GPU). This work demonstrates the tremendous potential of M3D-LIME with RRAM-based hybrid memory architecture for future data-centric applications.