One-step Computation Research Articles

Programmable In-Memory Computing Circuit for Solving Combinatorial Matrix Operation in One Step

Matrix operations are widely used in practical engineering, but the traditional processing methods rely on the loop iterations and neural network algorithm on the software, requiring a long time to calculate. To address such problem, this paper proposes full hardware in-memory computing circuits based on programmable memristor unit array that can solve combinatorial matrix operations of any order in just one step. First, two basic circuit modules are introduced, which can respectively solve matrix multiplication and matrix equation. Further, the basic modules can be linked to solve combinatorial matrix operations with different forms. It’s worth noting that every module can parallel program the value of each memristor in the memristor unit array and complete one-step computation by hardware. Then, some matrix operations are given in the paper as examples to prove the high accuracy of proposed method, where the average accuracy rate achieves <inline-formula xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"> <tex-math notation="LaTeX">$99\%$</tex-math> </inline-formula> . The PSPICE simulation results demonstrate that the processing speed is improved enormously according to the comparison of hardware and software. Moreover, the proposed method has broad application prospect in practical engineering, such as using designed combinational circuit to solve domain shift problem in zero-shot learning, which greatly accelerates the training process of zero-shot learning.

UCSL: Toward Unsupervised Common Subspace Learning for Cross-Modal Image Classification

The emerging research line of cross-modal learning focuses on the issue of transferring feature representation manner learned from limited multimodal data with labelings to the testing phase with partial modalities. This is essentially common and practical in the remote sensing community when only modal-incomplete data are in users’ hands due to inevitable imaging or access restrictions under large-scale observation scenarios. However, most of the existing cross-modal learning methods have been designed with exclusive reliance on labeling, which can be either limited or noisy due to their costly production. To address this issue, we explore in this paper the possibility to learn cross-modal feature representation in an unsupervised fashion. By integrating the multimodal data into a fully recombined matrix form, we propose 1) the use of common subspace representation as the regression target instead of conventionally adopted binary labels, and 2) the orthogonality and manifold alignment regularization terms to shrink the solution space whilst preserving the pairwise manifold correlations. Through this manner, the modality-specific and mutual latent representations in this common subspace as well as their corresponding projections can be learned simultaneously and their optimums can be efficiently reached through a nearly one-step computation with the help of Eigen decomposition. Finally, we show the superiority of our method through extensive image classification experiments on three multimodal datasets with four remotely sensed modalities involved (i.e., hyperspectral, multispectral, synthetic aperture radar, and light detection and ranging data). The code and dataset will be made freely available at https://github.com/jingyao16/UCSL after a possible publication to encourage the reproduction of our method and further use.

KNN Classification with One-step Computation

KNN classification is an improvisational learning mode, in which they are carried out only when a test data is predicted that set a suitable K value and search the K nearest neighbors from the whole training sample space, referred them to the lazy part of KNN classification. This lazy part has been the bottleneck problem of applying KNN classification due to the complete search of K nearest neighbors. In this paper, a one-step computation is proposed to replace the lazy part of KNN classification. The one-step computation actually transforms the lazy part to a matrix computation as follows. Given a test data, training samples are first applied to fit the test data with the least squares loss function. And then, a relationship matrix is generated by weighting all training samples according to their influence on the test data. Finally, a group lasso is employed to perform sparse learning of the relationship matrix. In this way, setting K value and searching K nearest neighbors are both integrated to a unified computation. In addition, a new classification rule is proposed for improving the performance of one-step KNN classification. The proposed approach is experimentally evaluated, and demonstrated that the one-step KNN classification is efficient and promising

Quantum interference enables constant-time quantum information processing.

It is an open question how fast information processing can be performed and whether quantum effects can speed up the best existing solutions. Signal extraction, analysis, and compression in diagnostics, astronomy, chemistry, and broadcasting build on the discrete Fourier transform. It is implemented with the fast Fourier transform (FFT) algorithm that assumes a periodic input of specific lengths, which rarely holds true. A lesser-known transform, the Kravchuk-Fourier (KT), allows one to operate on finite strings of arbitrary length. It is of high demand in digital image processing and computer vision but features a prohibitive runtime. Here, we report a one-step computation of a fractional quantum KT. The quantum d-nary (qudit) architecture we use comprises only one gate and offers processing time independent of the input size. The gate may use a multiphoton Hong-Ou-Mandel effect. Existing quantum technologies may scale it up toward diverse applications.

An alternative method for computing thermal stress in asphalt mixture: the Laplace transformation

Low-temperature cracking is one of the most severe distresses for asphalt pavement experiencing severely cold weather conditions. Many road authorities recognise thermal stress as a crucial parameter for evaluating the low-temperature performance of asphalt pavement. Thermal stress is conventionally computed with a two-step approach where the relaxation modulus is derived from the experimental creep compliance after which the convolution integral is numerically solved. In this paper, a one-step computation solution based on Laplace transformation is proposed. Thermal stress and corresponding critical cracking temperature of a set of six mixtures are easily computed and graphically and statistically compared to the values obtained with the traditional approach. It is observed that the use of the Laplace-based method provides reliable and reasonably close results to those obtained with the conventional solution.

I&lt;inline-formula&gt;&lt;tex-math notation="LaTeX"&gt;$^2$&lt;/tex-math&gt;&lt;alternatives&gt; &lt;inline-graphic xlink:type="simple" xlink:href="zhang-ieq1-2397438.gif"/&gt;&lt;/alternatives&gt;&lt;/inline-formula&gt;MapReduce: Incremental MapReduce for Mining Evolving Big Data

As new data and updates are constantly arriving, the results of data mining applications become stale and obsolete over time. Incremental processing is a promising approach to refreshing mining results. It utilizes previously saved states to avoid the expense of re-computation from scratch. In this paper, we propose i2MapReduce, a novel incremental processing extension to MapReduce, the most widely used frameworkfor mining big data. Compared with the state-of-the-art work on Incoop, i2MapReduce (i) performs key-value pair level incremental processing rather than task level re-computation, (ii) supports not only one-step computation but also more sophisticated iterative computation, which is widely used in data mining applications, and (iii) incorporates a set of novel techniques to reduce I/O overhead for accessing preserved fine-grain computation states. We evaluate i2MapReduce using a one-step algorithm and four iterative algorithms with diverse computation characteristics. Experimental results on Amazon EC2 show significant performance improvements of i2MapReduce compared to both plain and iterative MapReduce performing re-computation.

Interference cancellation beamforming robust to pointing errors

The conventional Wiener-Hopf beamformer is subject to substantial performance degradation in the presence of steering vector pointing errors. By removing the effects of the desired signal, the modified Wiener-Hopf beamformer avoids this problem but allows cochannel interferences to pass through in order to maximise the signal-to-noise ratio. In this study, a novel array beamformer is proposed, which not only reduces the effect of pointing errors, but also asymptotically provides complete interference rejection. In particular, the proposed beamformer utilises a vector space projection method and employs a one-step computation for the desired signal power. Using this, the effects of the desired signal can be extracted to form the desired-signal-absent covariance matrix. Thus, a weight vector orthogonal with the interference subspace can be constructed. Numerical results demonstrate the superior performance of the proposed beamformer in the presence of pointing errors relative to other existing approaches such as 'diagonal loading', 'robust Capon' and 'signal subspace projection' beamformers.

Finite-Element Inductance Computation in 2-D Eddy-Current Systems Using Sensitivity Analysis

A new finite-element (FE) method for the determination of differential and apparent inductances of massive conductors is proposed using a sensitivity analysis approach. The new technique is a one-step computation procedure in which the definition of distribution vectors makes possible a unified treatment of solid and filamentary conductors. Since massive conductors are included in the formulation, eddy-current and proximity effects are properly considered. The inductance matrix expression, formally derived from the sensitivity analysis, requires only one evaluation of the Jacobian matrix of the FE system at the solution point. As a result, an efficient calculation can be performed using the same solver employed in the FE solution. The computation of a multiport equivalent of an induction machine has been used to show the power of the proposed method. Our FE and multiport network results have been validated using a well-known commercial FE package.

Motion tracking based on area and level set weighted centroid shifting

In this study, the authors propose a stable colour-based tracking algorithm based on a new representation of the target location: the area weighted mean of the centroids corresponding to each colour bin of the target. The target location is well discriminated, since the centroids contain spatial information on the distribution of the colours and are rather insensitive to the loss of pixels and change in the number of pixels. The area weighting takes care that the major colours are treated with more importance than the minor colours. Due to these properties, it is possible to track the target in difficult conditions such as low-frame-rate environment, severe partial occlusion and partial colour change environment. Furthermore, the target localisation can be achieved in a one-step computation, which makes the algorithm fast. The authors compare the stability of the proposed tracking scheme with the original mean shift based tracker, both mathematically and experimentally. They also propose a background feature elimination algorithm, which is based on the level set based bimodal segmentation. The level set based bimodal segmentation segments out the region with dominant background feature and thus increases the robustness of the scheme.

High-order time discretizations in seismic modeling

Seismic modeling plays an important role in explor-ation geophysics. High-order modeling schemes are in demand for practical reasons. In this context, I present three kinds of high-order time discretizations: Lax-Wendroff methods, Nyström methods, and splitting methods. Lax-Wendroff methods are based on the Taylor expansion and the replacement of high-order temporal derivatives by spatial derivatives, Nyström methods are simplified Runge-Kutta algorithms, and splitting methods comprise substeps for one-step computation. Based on these methods, three schemes with third-order and fourth-order accuracy in time and pseudospectral discretizations in space are presented. I also compare their accuracy, stability, and computational complexity, and discuss advantages and shortcomings of these algorithms. Numerical experiments show that the fourth-order Lax-Wendroff scheme is more efficient for short-time simulations while the fourth-order Nyström scheme and the third-order splitting scheme are more efficient for long-term computations.

オブザーバを用いた離散時間1型サーボ系の一設計法

We propose a design of discrete-time type 1 servosystems using output feedback with one-step computation delay. As a basic design, we consider an output feedback design using the state feedback design proposed by Mita and Mukaida and a filtering type Kalman filter. We show that this controller can be simplified by use of a prediction type Kalman filter without sacrificing achievable feedback properties. For the servosystem with the simplified controller, we obtain decomposed expressions for the sensitivity matrices at the input and output of the plant. These expressions contain the sensitivity matrix of the regulator or that of the Kalman filter used in the design and explicitly represent the effect of the integrator and the computation delay in the controller. Using these expressions, we show that the duality holds between the feedback properties at the input and those at the output of the plant. In addition, we discuss the sensitivity properties obtained by the well-known loop transfer recovery techniques. A numerical example is presented to illustrate the performance of the proposed design.