Neural Networks (NN) are a family of models for a broad range of emerging machine learning and pattern recondition applications. NN techniques are conventionally executed on general-purpose ...processors (such as CPU and GPGPU), which are usually not energy-efficient since they invest excessive hardware resources to flexibly support various workloads. Consequently, application-specific hardware accelerators for neural networks have been proposed recently to improve the energy-efficiency. However, such accelerators were designed for a small set of NN techniques sharing similar computational patterns, and they adopt complex and informative instructions (control signals) directly corresponding to high-level functional blocks of an NN (such as layers), or even an NN as a whole. Although straightforward and easy-to-implement for a limited set of similar NN techniques, the lack of agility in the instruction set prevents such accelerator designs from supporting a variety of different NN techniques with sufficient flexibility and efficiency. In this paper, we propose a novel domain-specific Instruction Set Architecture (ISA) for NN accelerators, called Cambricon, which is a load-store architecture that integrates scalar, vector, matrix, logical, data transfer, and control instructions, based on a comprehensive analysis of existing NN techniques. Our evaluation over a total of ten representative yet distinct NN techniques have demonstrated that Cambricon exhibits strong descriptive capacity over a broad range of NN techniques, and provides higher code density than general-purpose ISAs such as ×86, MIPS, and GPGPU. Compared to the latest state-of-the-art NN accelerator design DaDianNao 5 (which can only accommodate 3 types of NN techniques), our Cambricon-based accelerator prototype implemented in TSMC 65nm technology incurs only negligible latency/power/area overheads, with a versatile coverage of 10 different NN benchmarks.
Effective and safe hemodialysis is essential for patients with acute kidney injury and chronic renal failures. However, the development of effective anticoagulant agents with safe antidotes for use ...during hemodialysis has proven challenging. Here, we describe DNA origami-based assemblies that enable the inhibition of thrombin activity and thrombus formation. Two different thrombin-binding aptamers decorated DNA origami initiates protein recognition and inhibition, exhibiting enhanced anticoagulation in human plasma, fresh whole blood and a murine model. In a dialyzer-containing extracorporeal circuit that mimicked clinical hemodialysis, the origami-based aptamer nanoarray effectively prevented thrombosis formation. Oligonucleotides containing sequences complementary to the thrombin-binding aptamers can efficiently neutralize the anticoagulant effects. The nanoarray is safe and immunologically inert in healthy mice, eliciting no detectable changes in liver and kidney functions or serum cytokine concentration. This DNA origami-based nanoagent represents a promising anticoagulant platform for the hemodialysis treatment of renal diseases.
DaDianNao: A Machine-Learning Supercomputer Yunji Chen; Tao Luo; Shaoli Liu ...
2014 47th Annual IEEE/ACM International Symposium on Microarchitecture,
2014-Dec.
Conference Proceeding
Many companies are deploying services, either for consumers or industry, which are largely based on machine-learning algorithms for sophisticated processing of large amounts of data. The ...state-of-the-art and most popular such machine-learning algorithms are Convolutional and Deep Neural Networks (CNNs and DNNs), which are known to be both computationally and memory intensive. A number of neural network accelerators have been recently proposed which can offer high computational capacity/area ratio, but which remain hampered by memory accesses. However, unlike the memory wall faced by processors on general-purpose workloads, the CNNs and DNNs memory footprint, while large, is not beyond the capability of the on chip storage of a multi-chip system. This property, combined with the CNN/DNN algorithmic characteristics, can lead to high internal bandwidth and low external communications, which can in turn enable high-degree parallelism at a reasonable area cost. In this article, we introduce a custom multi-chip machine-learning architecture along those lines. We show that, on a subset of the largest known neural network layers, it is possible to achieve a speedup of 450.65x over a GPU, and reduce the energy by 150.31x on average for a 64-chip system. We implement the node down to the place and route at 28nm, containing a combination of custom storage and computational units, with industry-grade interconnects.
Strict Watson–Crick base pairing and availability by automated synthesis have allowed deoxyribonucleic acid (DNA) molecules to be used as engineerable building blocks for constructing versatile ...nanostructures. In recent decades, with the development of DNA nanotechnology, a range of DNA‐based dynamic molecular devices with sophisticated nanostructures have been designed and constructed. Featuring programmability and biocompatibility, the applications of DNA‐based nanodevices have been extensively focused on the interfaces of biological systems. This review summarizes the recent progress in the design of DNA devices exhibiting programmable functions for biomedical applications. In vitro and in vivo applications of DNA‐based nanodevices in cellular imaging and systemic drug delivery are highlighted. The challenges and open opportunities are also discussed.
DNA‐based robotic devices have been designed and constructed in recent decades. Featuring programmability and biocompatibility, the applications of DNA devices have been extensively focused on the interfaces of biomedical systems. In this review, the recent developments of DNA devices working in cellular environments and animal levels are summarized.
The CRISPR/Cas9 gene-editing system has become a promising strategy for tumor therapy with its powerful oncogene-editing ability. However, the efficient delivery of sgRNA/Cas9 complex into target ...tumor cells remains a challenge. Herein, we report a facile strategy for the construction of an sgRNA/Cas9 complex co-assembled nanoplatform for targeted gene editing and combined tumor therapy. In our design, the TAT peptide and thiolated DNA linker functionalized gold nanorod can efficiently load the sgRNA/Cas9 complex through the hybridization between the 3′ overhang of sgRNA and the DNA linker. Due to the integration of an active cell targeting group (aptamer) and nuclear targeting peptide (TAT), the multifunctional nanoplatform can elicit the targeted cellular internalization and efficient nuclear targeting transportation to realize endogenous RNase H activated gene editing of the tumor-associated gene polo-like kinase 1 (PLK1). With mild photothermal treatment, this sgRNA/Cas9 complex loaded nanoplatform achieved efficient inhibition of tumor cell proliferation. This multifunctional nanocarrier provides a new strategy for the development of combined tumor therapy.
•A new reconstruction method for complex pipeline inspection is presented.•A centerline matching and registration algorithm for branched pipeline is proposed.•A pose estimation algorithm based on ...shape aspect graphs for joint is proposed.•A pipeline multi-camera measurement system is established to achieve our method.•The method’s efficiency and accuracy is higher than the present methods.
A fast and precise 3D reconstruction method for complex pipeline inspection is presented. The reconstruction process is divided into three phases based on multi-vision. A centerline matching and registration algorithm is proposed to achieve the branched and bended pipes’ reconstruction. The projected centerline of each pipe is split into several NURBS curves, and these curves are taken as the primitives of matching to be reconstructed and registered precisely. A pose estimation algorithm based on shape aspect graphs is proposed to achieve the joint’s reconstruction. The CAD model of each joint is used to obtain its shape aspect graphs as the templates which its edges image matches with. A pipeline multi-camera reconstruction system is established, and validation experiments are carried out. The results show that the proposed method completes a pipeline’s reconstruction within 5min. The reconstruction errors of its pipes and joints are 0.24mm and 0.496mm respectively.
•Partial expansion was observed in layers during foundation pit dewatering.•Model tests were performed to investigate the phenomenon suspected as an error.•Underlying confined aquifer was generated ...as a rigid plate with a hinge at one end.•Overlying layers were simulated with similitude materials.•Partial expansion occurred under coordination deformation and consolidation condition.
Partial expansion was observed in stratified subsidence during foundation pit dewatering. However, the phenomenon was suspected to be an error because the compression of layers is known to occur when subsidence occurs. A slice of the subsidence cone induced by drawdown was selected as the prototype. Model tests were performed to investigate the phenomenon. The underlying confined aquifer was generated as a movable rigid plate with a hinge at one end. The overlying layers were simulated with remolded materials collected from a construction site. Model tests performed under the conceptual model indicated that partial expansion occurred in stratified settlements under coordination deformation and consolidation conditions. During foundation pit dewatering, rapid drawdown resulted in rapid subsidence in the dewatered confined aquifer. The rapidly subsiding confined aquifer top was the bottom deformation boundary of the overlying layers. Non-coordination deformation was observed at the top and bottom of the subsiding overlying layers. The subsidence of overlying layers was larger at the bottom than at the top. The layers expanded and became thicker. The phenomenon was verified using numerical simulation method based on finite difference method. Compared with numerical simulation results, the boundary effect of the physical tests was obvious in the observation point close to the movable endpoint. The tensile stress of the overlying soil layers induced by the underlying settlement of dewatered confined aquifer contributed to the expansion phenomenon. The partial expansion of overlying soil layers was defined as inversed rebound. The inversed rebound was induced by inversed coordination deformation. Compression was induced by the consolidation in the overlying soil layers because of drainage. Partial expansion occurred when the expansion exceeded the compression. Considering the inversed rebound, traditional layer-wise summation method for calculating subsidence should be revised and improved.
Heavy metals are a growing threat to human health due to the resulting damage to the ecology; the removal of heavy metals by lactic acid bacteria (LAB) has been a focus of many studies. In this ...study, 10 LAB strains were evaluated for their ability to absorb and tolerate lead. Lactobacillus plantarum YW11 was found to possess the strongest ability of lead absorbing and tolerance, with the rate of absorption as high as 99.9% and the minimum inhibitory concentration of lead on YW11 higher than 1000 mg/L. Based on the isobaric tags for relative and absolute quantitation (iTRAQ) proteomics analysis of YW11, a total of 2009 proteins were identified both in the lead-treated strain and the control without the lead treatment. Among these proteins, 44 different proteins were identified. The abundance of 25 proteins increased significantly, and 19 proteins decreased significantly in the treatment group. These significantly differential abundant proteins are involved in the biological processes of amino acid and lipid metabolism, energy metabolism, cell wall biosynthesis, and substance transport. This study contributed further understanding of the molecular mechanism of L. plantarum in the binding and removal of lead to explore its potential application in counteracting heavy metal pollution of environment, food, and other fields.
9-
-epoxycarotenoid dioxygenase (NCED) is a key enzyme involved in the biosynthesis of abscisic acid (ABA), which is associated with drought tolerance in plants. An osmotic-inducible
gene was ...isolated from a drought-resistant cultivar of
and constitutively overexpressed in a drought-sensitive cultivar of
. Transgenic plants showed significantly improved drought tolerance, including a higher growth rate and better drought resistant under drought conditions, compared to those of wild-type (WT) plants. After water was withheld for 50 days, the upper leaves of transgenic plants remained green, whereas most leaves of WT plants turned yellow and fell. Besides the increase in ABA content, overexpression of
induced the production of jasmonic acid (JA) and accumulation of JA biosynthesis-related genes, including
(
) and
(
). Moreover, transgenic plants possessed advantageous physiological indices, including lower leaf stomatal density, lower photosynthesis rate, and lower accumulation of proline and superoxide dismutase (SOD), compared to those of WT plants, indicating increased resistance to drought stress. Quantitative real time polymerase chain reaction (RT-qPCR) analysis revealed that overexpression of
enhanced the expression of drought-responsive genes, such as
(
)
(
),
(
),
(
) and
(
). Although the development of transgenic plants was delayed by 4 months than WT plants, because of seed dormancy and abnormal seedlings, the surviving transgenic plants provided a solid method for protection of woody plants from drought stress.
To understand the molecular mechanism involved in the survivability of cold-tolerant lactic acid bacteria was of great significance in food processing, since these bacteria play a key role in a ...variety of low-temperature fermented foods. In this study, the cold-stress response of probiotic
K25 isolated from Tibetan kefir grains was analyzed by iTRAQ proteomic method. By comparing differentially expressed (DE) protein profiles of the strain incubated at 10°C and 37°C, 506 DE proteins were identified. The DE proteins involved in carbohydrate, amino acid and fatty acid biosynthesis and metabolism were significantly down-regulated, leading to a specific energy conservation survival mode. The DE proteins related to DNA repair, transcription and translation were up-regulated, implicating change of gene expression and more protein biosynthesis needed in response to cold stress. In addition, two-component system, quorum sensing and ABC (ATP-binding cassette) transporters also participated in cell cold-adaptation process. These findings provide novel insight into the cold-resistance mechanism in
with potential application in low temperature fermented or preserved foods.