Showing papers by "Dong Wang published in 2015"

Complex Learning in Bio-plausible Memristive Networks

Abstract: The emerging memristor-based neuromorphic engineering promises an efficient computing paradigm. However, the lack of both internal dynamics in the previous feedforward memristive networks and efficient learning algorithms in recurrent networks, fundamentally limits the learning ability of existing systems. In this work, we propose a framework to support complex learning functions by introducing dedicated learning algorithms to a bio-plausible recurrent memristive network with internal dynamics. We fabricate iron oxide memristor-based synapses, with well controllable plasticity and a wide dynamic range of excitatory/inhibitory connection weights, to build the network. To adaptively modify the synaptic weights, the comprehensive recursive least-squares (RLS) learning algorithm is introduced. Based on the proposed framework, the learning of various timing patterns and a complex spatiotemporal pattern of human motor is demonstrated. This work paves a new way to explore the brain-inspired complex learning in neuromorphic systems.

Development of a neuromorphic computing system

Abstract: Although a variety of solutions for neuromorphic systems based on different hardware technology and software programming schemes, there has yet to be a common accepted one. Based on some recent findings in brain science, we propose a new design rule for developing a brain inspired computing system. We design and fabricate a neuromorphic chip, named ‘Tianji’ chip. A multi-chip architecture-based PCB board has been demonstrated. The detailed hardware implementation and software programming scheme are presented in this paper.

Tissue-specific mechanical and geometrical control of cell viability and actin cytoskeleton alignment.

Abstract: Different tissues have specific mechanical properties and cells of different geometries, such as elongated muscle cells and polygonal endothelial cells, which are precisely regulated during embryo development. However, the mechanisms that underlie these processes are not clear. Here, we built an in vitro model to mimic the cellular microenvironment of muscle by combining both mechanical stretch and geometrical control. We found that mechanical stretch was a key factor that determined the optimal geometry of myoblast C2C12 cells under stretch, whereas vascular endothelial cells and fibroblasts had no such dependency. We presented the first experimental evidence that can explain why myoblasts are destined to take the elongated geometry so as to survive and maintain parallel actin filaments along the stretching direction. The study is not only meaningful for the research on myogenesis but also has potential application in regenerative medicine.

Generating intense fully coherent soft x-ray radiation based on a laser-plasma accelerator

Abstract: Laser-plasma based accelerator has the potential to dramatically reduce the size and cost of future x-ray light sources to the university-laboratory scale. However, the large energy spread of the laser-plasma accelerated electron beam may hinder the way for short wavelength free-electron laser generation. In this paper, we propose a novel method for directly imprinting strong coherent micro-bunching on the electron beam with large intrinsic energy spread by using a wavefront-tilted conventional optical laser beam and a weak dipole magnet. Theoretical analysis and numerical simulations demonstrate that this technique can be used for the generation of fully coherent femtosecond soft x-ray radiation at gigawatts level with a very short undulator.

Balancing the Expression and Production of a Heterodimeric Protein: Recombinant Agkisacutacin as a Novel Antithrombotic Drug Candidate

Abstract: Balancing the Expression and Production of a Heterodimeric Protein: Recombinant Agkisacutacin as a Novel Antithrombotic Drug Candidate

Excellent performances of energy harvester using cantilever driving double-clamped 0.7Pb(Mg1/3Nb2/3)O3-0.3PbTiO3 plates and symmetric middle-stops

Abstract: We present a high performance nonlinear piezoelectric energy harvester constituted by a cantilever with symmetrically middle-stops and double-clamped piezoelectric plates based on piezoelectric single crystal 0.7Pb(Mg1/3Nb2/3)O3-0.3PbTiO3. Electrical properties of the device under different excitation frequencies, accelerations, and load resistances are studied systematically. Under a low acceleration of 3 m/s2 (0.3 g), a peak voltage of 26.2 V and a maximum normalized power of 25.6 mW/g2 were obtained across a matching impedance of 600 kΩ with favorable bandwidths. The low excitation acceleration and excellent performances indicate that the device can be a promising candidate for energy harvesting in low-power electronics and wireless sensors.

Micro-Raman spectroscopy characterization of silicon with different structures irradiated with energetic Bi-ions

Abstract: Researches of irradiation effects on silicon possess not only fundamental interests but also potential application prospects. Comparison studies about structural modification of silicon materials with different structures under identical irradiation conditions can reveal the irradiation mechanisms for amorphous and crystalline phases of silicon. For this purpose, amorphous silicon (a-Si) and nano-crystalline silicon (nc-Si) films as well as mono-crystalline silicon (c-Si) samples were irradiated with 6.0 MeV Bi-ions at room temperature. The ion fluences are 1.0 × 10 13 , 5.0 × 10 13 and 1.0 × 10 14 ions/cm 2 . All samples were analyzed by using a Raman spectrometer. The obtained results show that the crystalline fraction of c-Si and nc-Si decrease with increasing fluence, which indicates that the irradiation induces the amorphization of nc-Si and c-Si samples. In addition, the variation in Raman frequency of crystalline peak after irradiation reveals that the irradiation also results in the increased stress in crystalline phase of c-Si and nc-Si samples. As the fluence increases, the bond angle deviation and the ratio of TA to TO mode of amorphous network of a-Si and nc-Si films initially increase and then decrease by a diminishing degree, while the bond angle deviation and the ratio of TA to TO mode of amorphous network of c-Si samples increase continuously. This gives the dependence of short-range structural order of amorphous network of a-Si, nc-Si and c-Si samples on the ion fluence, which is related with the irradiation induced variation of local free energy. It is considered that the irradiation induced structural modification of silicon samples is mainly attributed to the nuclear energy loss. The irradiation effects of energetic heavy-ions on crystalline and amorphous phases of silicon have been discussed, respectively.

TEM Characterization of Helium Bubbles in T91 and MNHS Steels Implanted with 200 keV He Ions at Different Temperatures

Abstract: Modified novel high silicon steel (MNHS, a newly developed reduced-activation martensitic alloy) and commercial alloy T91 are implanted with 200 keV He2+ ions to a dose of 5 × 1020 ions/m2 at 300, 450 and 550°C. Transmission electron microscopy (TEM) is used to characterize the size and morphology of He bubbles. With the increase of the implantation temperature, TEM observations indicate that bubbles increase in size and the proportion of 'brick shaped' cuboid bubbles increases while the proportion of polyhedral bubbles decreases in both the steel samples. For the samples implanted at the same temperature, the average size of He bubbles in MNHS is smaller than that in T91. This might be due to the abundance of boundaries and precipitates in MNHS, which provide additional sites for the trapping of He atoms, thus reduce the susceptibility of MNHS to He embrittlement.

FPGA-based neuromorphic computing system with a scalable routing network

Abstract: In this paper, the design and FPGA implementation of a multi-core neuromorphic computing system is presented. This system consists of several computing cores that are connected through an on-chip routing network. Each core is capable of the computation of a 2-layer neural network with a variable number of axon inputs and neurons. This system was built on an Altera FPGA and then examined with a 3-layer RBM and a 5-layer DBN for handwritten digits recognition. Compared to an Intel Core i3 CPU, the multi-core system achieves higher computing speed despite of tiny loss of accuracy.

Universal Controlled-Phase Gates Between Distant Atoms Separately Trapped in Thermal Cavities

Abstract: Nonlocal implementation of a family of universal two-qubit controlled-phase gates is investigated in the regime of cavity quantum electrodynamics (CQED) system. With the driving of classical field, the atoms taken as qubits can be efficiently coupled to thermal cavities. In the course of the implementation, atomic spontaneous emission and cavity decay are negligible. With the help of quantum repeaters, our proposal can accomplish entanglement-based a family of universal controlled-phase logic gates for remote two atoms with high fidelity and success probability. Remarkably, our proposal is able to break through the limitation that error probability scales exponentially with the length of the channel. Furthermore, experimental feasibility of the scenario presented is evaluated and it turns out that it is realizable according to today’s CQED technologies. It is here emphasised that our proposal might be important to long-distance communication in prospective quantum multi-node networks.

Extensible neuromorphic computing simulator based on a programmable hardware

Abstract: The CPU based software simulator for neuromorphic computing is faced with problems of high power consumption and speed limitation of interconnection network, especially in large-scale Spiking Neural Network (SNN) simulation. IBM, Stanford, and ARM have demonstrated us their solutions for neuromorphic computing. However, the cost and development cycle make these approaches impractical for general research. In this paper, we provide an extensible and programmable hardware platform which is suitable for large-scale neuromorphic computing and simulation. The design of hardware simulator is based on Altera Stratix V FPGA, and its most significant advantage consists in parallel processing capability. With NIOS II soft core processor and Gigabit Ethernet interface, the simulator is programmable for different applications and is very adaptable for neural network configuration. Two kinds of routing topologies are available for network performance simulation. By inserting statistic modules into the computing core, it's possible to monitor the single core state as well as the network state, which is of great importance in routing strategy establishing. Furthermore, this extensible simulator supports at least 105-neuron-network computing.