Search

The authors have numerically studied how to enhance reservoir computing performance by thoroughly extracting their spin-wave device potential for higher-dimensional information generation. The reservoir device has a 1-input exciter and 120-output detectors on the top of a continuous magnetic garnet film for spin-wave transmission. For various nonlinear and fading-memory dynamic phenomena distributing in the film space, small in-plane magnetic fields were used to prepare stripe domain structures and various damping constants at the film sides and bottom were explored. The ferromagnetic resonant frequency and relaxation time of spin precession clearly characterized the change in spin dynamics with the magnetic field and damping constant. The common input signal for reservoir computing was a 1 GHz cosine wave with random 6-valued amplitude modulation. A basic 120-dimensional reservoir output vector was obtained from time-series signals at the 120 output detectors under each of the three magnetic field conditions. Then, 240- and 360-dimensional reservoir output vectors were also constructed by concatenating two and three basic ones, respectively. In nonlinear autoregressive moving average (NARMA) prediction tasks, the computational performance was enhanced as the dimension of the reservoir output vector becomes higher and a significantly low prediction error was achieved for the 10th-order NARMA using the 360-dimensional vector and optimum damping constant. The results are clear evidence that the collection of diverse output signals efficiently increases the dimensionality effective for reservoir computing, i.e., reservoir-state richness. This paper demonstrates that performance enhancement through various configuration settings is a practical approach for on-chip reservoir computing devices with small numbers of real output nodes.

This paper proposes a novel method of independent component analysis (ICA), which we name higher-order tensor ICA (HOT-ICA). HOT-ICA is a tensor ICA that makes effective use of the signal categories represented by the axes of a separating tensor. Conventional tensor ICAs, such as multilinear ICA (MICA) based on Tucker decomposition, do not fully utilize the high dimensionality of tensors because the matricization in MICA nullifies the tensor axial categorization. In this paper, we deal with multiple-target signal separation in a multiple-input multiple-output (MIMO) radar system to detect respiration and heartbeat. HOT-ICA realizes high robustness in learning by incorporating path information, i.e., the physical-measurement categories on which transmitting/receiving antennas were used. In numerical-physical experiments, our HOT-ICA system effectively separate the bio-signals successfully even in an obstacle-affecting environment, which is usually a difficult task. The results demonstrate the significance of the HOT-ICA, which keeps the tensor categorization unchanged for full utilization of the high-dimensionality of the separation tensor.

This paper studies numerically how the signal detector arrangement influences the performance of reservoir computing using spin waves excited in a ferrimagnetic garnet film. This investigation is essentially important since the input information is not only conveyed but also transformed by the spin waves into high-dimensional information space when the waves propagate in the film in a spatially distributed manner. This spatiotemporal dynamics realizes a rich reservoir-computational functionality. First, we simulate spin waves in a rectangular garnet film with two input electrodes to obtain spatial distributions of the reservoir states in response to input signals, which are represented as spin vectors and used for a machine-learning waveform classification task. The detected reservoir states are combined through readout connection weights to generate a final output. We visualize the spatial distribution of the weights after training to discuss the number and positions of the output electrodes by arranging them at grid points, equiangularly circular points or at random. We evaluate the classification accuracy by changing the number of the output electrodes, and find that a high accuracy ($>$ 90\%) is achieved with only several tens of output electrodes regardless of grid, circular or random arrangement. These results suggest that the spin waves possess sufficiently complex and rich dynamics for this type of tasks. Then we investigate in which area useful information is distributed more by arranging the electrodes locally on the chip. Finally, we show that this device has generalization ability for input wave-signal frequency in a certain frequency range. These results will lead to practical design of spin-wave reservoir devices for low-power intelligent computing in the near future.

Synthetic aperture radar (SAR) is widely used for ground surface classification since it utilizes information on vegetation and soil unavailable in optical observation. Image classification often employs convolutional neural networks. However, they have serious problems such as long learning time and resolution degradation in their convolution and pooling processes. In this paper, we propose complex-valued reservoir computing (CVRC) to deal with complex-valued images in interferometric SAR (InSAR). We classify InSAR image data by using CVRC successfully with a higher resolution and a lower computational cost, i.e., one-hundredth learning time and one-fifth classification time, than convolutional neural networks. We also conduct experiments on slope angle estimation. CVRC is found applicable to quantitative tasks dealing with continuous values as well as discrete classification tasks with a higher accuracy., Comment: 35 pages, 16 figures

Spin waves propagating through a stripe domain structure and reservoir computing with their spin dynamics have been numerically studied with focusing on the relation between physical phenomena and computing capabilities. Our system utilizes a spin-wave-based device that has a continuous magnetic garnet film and 1-input/72-output electrodes on the top. To control spatially-distributed spin dynamics, a stripe magnetic domain structure and amplitude-modulated triangular input waves were used. The spatially-arranged electrodes detected spin vector outputs with various nonlinear characteristics that were leveraged for reservoir computing. By moderately suppressing nonlinear phenomena, our system achieves 100$\%$ prediction accuracy in temporal exclusive-OR (XOR) problems with a delay step up to 5. At the same time, it shows perfect inference in delay tasks with a delay step more than 7 and its memory capacity has a maximum value of 21. This study demonstrated that our spin-wave-based reservoir computing has a high potential for edge-computing applications and also can offer a rich opportunity for further understanding of the underlying nonlinear physics., Comment: 14 pages, 13 figures

A pulsed oscillating power amplifier has been developed for high frequency biasing\cite{kn:deb1} and real time turbulent feedback experiment in STOR-M tokamak. It is capable to provide output peak to peak oscillating voltage of around $\pm60$V and current around 30A within frequency band 1kHz-50kHz without any distortion of any waveform signal. Overall output power is amplified by two stages power mosfet op-amp as well as nine identical push-pull amplifiers which are parallel connected in final stages. The power amplifier input signal, collected from plasma floating potential during plasma shot, is optically isolated with tokamak vessel for real time feedback experiment. Here, filtered floating potential fluctuations having band width between 5kHz-40kHz has been amplified and fed to an electrode inserted into the plasma edge to study response of plasma turbulence. It is observed that magnetic fluctuations are suppressed due to real time feedback of floating potential., Comment: No

A pulsed oscillating power amplifier has been developed to apply high frequency biasing voltage to an electrode at the edge of the STOR-M tokamak plasma. The power amplifier can deliver a peak-to-peak oscillating voltage up to 120V and current 30A within the frequency range of 1kHz-50kHz. The electrode is located in the equatorial plane at radius $\rho = 0.88$. The frequency of the applied voltage has been varied between discharges. It is observed that the plasma density and soft x-ray intensity from the plasma core region usually increase at lower frequency regime 1kHz-5kHz as well as relatively higher frequency regime 20kHz-25kHz but seldom increase in between them. Increment of $\tau_{p}$ \& $\tau_{E}$ have been observed from the derivations of experimental data in both frequency regimes. Transport simulation has been carried out using the ASTRA simulation code for STOR-M tokamak parameters to understand the physical process behind experimental observations at higher frequency branch. The model is based on GAM excitement at resonance frequency associated with Ware-pinch due to oscillating electric field produced by biasing voltage which can suppress anomalous transport. Simulation results reproduce the experiment quite well in terms of the density, particle confinement as well as energy confinement time evolution. All those results indicate high frequency biasing is capable of improving confinement efficiently.

Injection of relatively high density spheromaks with significant helicity-content into a tokamak has been proposed as a means for fueling and current drive. The CHI (Co-axial Helicity Injection) device was devised to inject current to the STOR-M tokamak. Various circuit modifications were made to the CHI controls, enabling testing of various injection configurations. The charge/discharge circuits for CT formation/acceleration and stuffing field were modified, and the power supplies and power converters were replaced. Various modifications were implemented to solve the original slow bank triggering problems. The CHI device was mounted on STOR-M for radial and vertical CT injection at various times. Spheromak injection into STOR-M usually resulted in disruption of the tokamak discharge. After modifying the CHI device to operate at increased power, it looked like tokamak current was increased by a few kiloamps just prior to disruption, but careful testing proved that the signals indicating a current drive were actually spurious, caused by inductive pickup. The CHI device was attached to a portable vacuum chamber that was constructed from spare parts, to characterise the CTs produced. Magnetic probes were constructed to measure poloidal and toroidal field near the CT edge. Langmuir probes were made and returned reasonable estimates for edge CT density and temperature., Comment: 22 pages, 17 figures

A model for plasma/neutral fluid interaction was developed and included in the DELiTE code framework implementation of non-linear MHD equations. The source rates of ion, electron and neutral fluid momentum and energy due to ionization and recombination are derived using a simple method that enables determination of the volumetric rate of thermal energy transfer from electrons to photons and neutral particles in the radiative recombination reaction. This quantity can not be evaluated with the standard formal procedure of taking moments of the relevant collision operator, and has been neglected in other studies. The plasma/neutral fluid interaction model was applied to study CT formation in the SMRT and SPECTOR magnetized Marshall guns, enabling clarification of the mechanisms behind the significant increases in CT electron density that are routinely observed well after formation on the SPECTOR experiment. Neutral gas, which remains concentrated below the gas valves after CT formation, diffuses up the gun barrel to the CT containment region where it is ionized, leading to the observed electron density increases. This understanding helps account for the exceptionally significant increase in temperature, and markedly reduced density, observed during the electrode edge biasing experiment conducted on SPECTOR. It is thought that edge fueling impediment, a consequence of a biasing-induced transport barrier, is largely responsible for the observed temperature increase and density decrease., Comment: 26 pages, 32 figures

A description of an edge-biasing experiment conducted on the SPECTOR (Spherical Compact Toroid) plasma injector is presented, along with initial results. The insertion of a disc-shaped molybdenum electrode (probe), biased at up to +100V, into the edge of the CT (Compact Torus), resulted in up to 1kA radial current being drawn. Core electron temperature, as measured with a Thomson-scattering diagnostic, was found to increase by a factor of up to 2.4 in the optimal configuration tested. H_alpha intensity was observed to decrease, and CT lifetimes increased by a factor of up to 2.3. A significant reduction in electron density was observed; this is thought to be due to the effect of a transport barrier impeding CT fueling, where, as verified by MHD simulation, the fueling source is neutral gas that remains concentrated around the gas valves after CT formation., Comment: 15 pages, 13 figures

The magnetic compression experiment at General Fusion was a repetitive non-destructive test to study plasma physics to Magnetic Target Fusion compression. A compact torus (CT) is formed with a co-axial gun into a containment region with an hour-glass shaped inner flux conserver, and an insulating outer wall. External coil currents keep the CT off the outer wall (radial levitation) and then rapidly compress it inwards. The optimal external coil configuration greatly improved both the levitated CT lifetime and the rate of shots with good flux conservation during compression. As confirmed by spectrometer data, the improved levitation field profile reduced plasma impurity levels by suppressing the interaction between plasma and the insulating outer wall during the formation process. Significant increases in magnetic field, density, and ion temperature were routinely observed at magnetic compression despite the prevalence of an instability, thought be an external kink, at compression. Matching the decay rate of the levitation coil currents to that of the internal CT currents resulted in a reduced level of MHD activity associated with unintentional compression by the levitation field, and a higher probability of long-lived CTs. An axisymmetric finite element MHD code that conserves system energy, particle count, angular momentum, and toroidal flux, was developed to study CT formation into a levitation field and magnetic compression. An overview of the principal experimental observations, and comparisons between simulated and experimental diagnostics are presented., Comment: 43 pages, 98 figures. Partially presented in conference posters C. Dunlea et al., Magnetic Compression at General Fusion - Experiment & Simulation with Neutral Fluid, APS DPP Milwaukee 2017, EPS Prague 2018, and ICPP Vancouver 2018

This paper proposes online-learning complex-valued neural networks (CVNNs) to predict future channel states in fast-fading multipath mobile communications. CVNN is suitable for dealing with a fading communication channel as a single complex-valued entity. This framework makes it possible to realize accurate channel prediction by utilizing its high generalization ability in the complex domain. However, actual communication environments are marked by rapid and irregular changes, thus causing fluctuation of communication channel states. Hence, an empirically selected stationary network gives only limited prediction accuracy. In this paper, we introduce regularization in the update of the CVNN weights to develop online dynamics that can self-optimize its effective network size by responding to such channel-state changes. It realizes online adaptive, highly accurate and robust channel prediction with dynamical adjustment of the network size. We demonstrate its online adaptability in simulations and real wireless-propagation experiments., Comment: 40 pages, 9 figures

Reservoir computing is a computational framework suited for temporal/sequential data processing. It is derived from several recurrent neural network models, including echo state networks and liquid state machines. A reservoir computing system consists of a reservoir for mapping inputs into a high-dimensional space and a readout for pattern analysis from the high-dimensional states in the reservoir. The reservoir is fixed and only the readout is trained with a simple method such as linear regression and classification. Thus, the major advantage of reservoir computing compared to other recurrent neural networks is fast learning, resulting in low training cost. Another advantage is that the reservoir without adaptive updating is amenable to hardware implementation using a variety of physical systems, substrates, and devices. In fact, such physical reservoir computing has attracted increasing attention in diverse fields of research. The purpose of this review is to provide an overview of recent advances in physical reservoir computing by classifying them according to the type of the reservoir. We discuss the current issues and perspectives related to physical reservoir computing, in order to further expand its practical applications and develop next-generation machine learning systems., Comment: 62 pages, 13 figures