Small Kernel Research Articles

Hysteretic behavior of spatially coupled phase-oscillators

Motivated by phenomena related to biological systems such as the synchronously flashing swarms of fireflies, we investigate a network of phase oscillators evolving under the generalized Kuramoto model with inertia. A distance-dependent, spatial coupling between the oscillators is considered. Zeroth and first order kernel functions with finite kernel radii were chosen to investigate the effect of local interactions. The hysteretic dynamics of the synchronization depending on the coupling parameter was analyzed for different kernel radii. Numerical investigations demonstrate that (1) locally locked clusters develop for small coupling strength values, (2) the hysteretic behavior vanishes for small kernel radii, (3) the ratio of the kernel radius and the maximal distance between the oscillators characterizes the behavior of the network.

Analysis of premixed flame kernel/turbulence interactions under engine conditions based on direct numerical simulation data

Although the evolution of premixed flames in turbulence has been frequently studied, it is not well understood how small flames interact with large-scale turbulent flow motion. Since this question is of practical importance for the occurrence of cycle-to-cycle variations in spark ignition engines, the objective of the present work is to fundamentally differentiate early flame kernel development from well-established turbulent flame configurations. For this purpose, a DNS database consisting of three flames propagating in homogeneous isotropic turbulence (Falkenstein et al., Combust. Flame, 2019) is considered. The flames feature different ratios of the initially laminar flame diameter to the integral length scale. To quantify flame kernel development, the time evolution of flame topology and flame front geometry are analysed in detail. It is shown that some realizations of the early flame kernel are substantially influenced by high compressive strain caused by large-scale turbulent flow motion with characteristic length scales greater than the flame kernel size. As a result, the initial spherical kernel topology may become highly distorted, which is reflected in the stochastic occurrence of excessive curvature variance. Two mechanisms of curvature production resulting from early flame kernel/turbulence interactions are identified by analysis of the mean curvature balance equation. Further, it is shown that the curvature distribution of small flame kernels becomes strongly skewed towards positive curvatures, which is contrary to developed turbulent flames. Hence, the transition of ignition kernels to self-sustaining turbulent flames is very different in nature compared with the development of a statistically planar flame brush.

Intelligent Fault Diagnosis for Rotating Machines Using Deep Learning

Abstract The diagnosis of failures in high-speed machining centers and other rotary machines is critical in manufacturing systems, because early detection can save a representative amount of time and cost. Fault diagnosis systems generally have two blocks: feature extraction and classification. Feature extraction affects the performance of the prediction model, and essential information is extracted by identifying high-level abstract and representative characteristics. Deep learning (DL) provides an effective way to extract the characteristics of raw data without prior knowledge, compared with traditional machine learning (ML) methods. A feature learning approach was applied using one-dimensional (1-D) convolutional neural networks (CNN) that works directly with raw vibration signals. The network structure consists of small convolutional kernels to perform a nonlinear mapping and extract features; the classifier is a softmax layer. The method has achieved satisfactory performance in terms of prediction accuracy that reaches ∼99 % and ∼97 % using a standard bearings database: the processing time is suitable for real-time applications with ∼8 ms per signal, and the repeatability has a low standard deviation <2 % and achieves an acceptable network generalization capability.

Technical Note: A dose calculation framework for dynamic electron arc radiotherapy (DEAR) using VirtuaLinac Monte Carlo simulation tool.

Dynamic electron arc radiotherapy (DEAR) is a novel dynamic technique that achieves highly conformal dose through simultaneous couch and gantry motion during delivery. The purpose of this study is to develop a framework integrating a Monte Carlo dose engine (VirtuaLinac) to a treatment planning system (TPS, Eclipse) for DEAR. A quality assurance (QA) procedure is also developed. The interfaces include the following: computed tomography image export and conversion for VirtuaLinac; VirtuaLinac computation tasks management through application programming interface (API); and dose matrix processing and evaluation. The framework was validated with both static beam and DEAR plan with a 3×3cm2 cutout for both 6 and 9MeV electrons. Verification plans for DEAR were created on flat phantom and a hybrid dose calculation technique was developed which convolves precalculated small field kernel with the beam trajectory, and the resulting dose was compared with the full VirtuaLinac calculation and film measurement. Excellent agreement between VirtuaLinac and eMC was observed with three-dimensional γ pass rate of 98% at 1%/1mm criteria for both 6 and 9MeV electrons. Film measurement shows two-dimensional (2D) γ passing rate of 99.8 % (6MeV) and 97.1% (9MeV) at 2%/2mm criteria. For DEAR plans the comparison of VirtuaLinac and measurement shows the 2D γ passing rates of 94% at 2%/2mm for 6MeV. The dose distributions from hybrid method in phantom are identical to the full VirtuaLinac simulations, but can be done instantly. A framework has been developed for DEAR dose calculation using VirtuaLinac Monte Carlo dose engine. The VirtuaLinac calculated dose was validated against measurement. A feasible and practical DEAR QA method has been developed for dose measurement in phantom. The hybrid dose calculation technique is efficient and suitable for DEAR QA purpose.

Handwriting posture prediction based on unsupervised model

Writing is an important basic skill for humans. To acquire such a skill, pupils often have to practice writing for several hours each day. However, different pupils usually possess distinct writing postures. Bad postures not only affect the speed and quality of writing, but also severely harm the healthy development of pupils’ spine and eyesight. Therefore, it is of key importance to identify or predict pupils’ writing postures and accordingly correct bad ones. In this paper, we formulate the problem of handwriting posture prediction for the first time. Further, we propose a neural network constructed with small convolution kernels to extract features from handwriting, and incorporate unsupervised learning and handwriting data analysis to predict writing postures. Extensive experiments reveal that our approach achieves an accuracy rate of 93.3%, which is significantly higher than the 76.67% accuracy of human experts.

Klasifikasi percepatan dari sinyal gempa bumi dan sinyal linier aktivitas manusia menggunakan akselerometer smartphone berbasis algoritme support vector machine

The threat of earthquake calamity spread throughout most of the Indonesian archipelago. Smartphone’s accelerometer usage as a seismic parameter detector in Indonesia, of which the noise has obstacles, mainly due to human activities. This study aims to classify linear acceleration signals caused by human activity and earthquake acceleration signals as an initial effort to reduce noise caused by human activity in the smartphone’s accelerometer signals. Both signals are classified by using the Support Vector Machine (SVM) algorithm of which consists of several steps, respectively, data collection, data preprocessing, data segmentation, feature extraction, and classification. These algorithms are tested to 2545 human activity signals in trouser pocket, 2430 human activity signals in shirt pocket and earthquake acceleration signals. Based on the test results by using the confusion matrix, linear acceleration signal data caused by human activity and earthquake acceleration signals can be classified properly using an SVM algorithm with Polynomial or Gaussian kernel with a small kernel scale value. The algorithms can achieve an accuracy of 87.74% to 97.94%.

SMALL KERNEL4is required for mitochondrial cox1 transcript editing and seed development in maize.

In land plants, cytidine-to-uridine (C-to-U) editing of organellar transcripts is an important post-transcriptional process, which is considered to remediate DNA genetic mutations to restore the coding of functional proteins. Pentatricopeptide repeat (PPR) proteins have key roles in C-to-U editing. Owing to its large number, however, the biological functions of many PPR proteins remain to be identified. Through characterizing a small kernel4 (smk4) mutant, here we report the function of Smk4 and its role in maize growth and development. Null mutation of Smk4 slows plant growth and development, causing small plants, delayed flowering time, and small kernels. Cloning revealed that Smk4 encodes a new E-subclass PPR protein, and localization indicated that SMK4 is exclusively localized in mitochondria. Loss of Smk4 function abolishes C-to-U editing at position 1489 of the cytochrome c oxidase1 (cox1) transcript, causing an amino acid change from serine to proline at 497 in Cox1. Cox1 is a core component of mitochondrial complex IV. Indeed, complex IV activity is reduced in the smk4, along with drastically elevated expression of alternative oxidases (AOX). These results indicate that SMK4 functions in the C-to-U editing of cox1-1489, and this editing is crucial for mitochondrial complex IV activity, plant growth, and kernel development in maize.

Extreme semi-supervised learning for multiclass classification

Semi-Supervised Support Vector Machines (S3VMs) provide a powerful framework for Semi-Supervised Learning (SSL) tasks which leverage widely available unlabeled data to improve performance. However, there exist three issues in S3VMs: (i) S3VMs require concurrently training c one-against-all (OAA) classifiers (c is the number of classes) for multiclass classification, which is prohibitive for large c; (ii) S3VMs require huge computational time and large storage (because of the large kernel matrix) in large-scale training and testing; (iii) S3VMs require the balance constraint in the unlabeled data, which not only needs prior knowledge from the unlabeled data (the prior knowledge is unavailable in some applications), but also makes their nonconvex optimization problem more intractable. To address these issues, a novel method called Extreme Semi-Supervised Learning (ESSL) is proposed in this paper. First, the framework of Extreme Learning Machine (ELM) is adopted to handle both binary and multiclass classification problems in a unified model. Second, the hidden layer is encoded by an extremely small approximate empirical kernel map (AEKM) to greatly reduce the computational cost and the memory usage for training and testing. Third, the balance constraint (prior knowledge) in the unlabeled data is removed through the elaborative design of weighting function (which emphasizes the importance of labeled data and the minority pattern in the labeled data).By these three ways, ESSL can be solved effectively and efficiently based on alternating optimization (AO). More specifically, ESSL can be analytically and simply solved by generalized pseudoinverse and oneHotMap function (without any optimization solver and the OAA strategy) in the AO procedure, and consequently, better performance and much faster training speed are always achieved in ESSL. Our empirical study shows that ESSL significantly outperforms existing efficient SSL methods (e.g., meanS3VM and SS-ELM) in terms of accuracy, efficiency and memory, especially for large-scale multiclass problems. As an example, on the 20Newsgroups dataset, ESSL respectively runs 45 and 120 times faster than meanS3VM for training and testing with the improvement in accuracy of 3%, while the memory usage is reduced to 1/14. It is noteworthy that even though all the model parameters are with default values, ESSL already produces very excellent performance without fine-tuning parameters.

Super-resolution reconstruction of knee magnetic resonance imaging based on deep learning

Background and objectiveWith the rapid development of medical imaging and intelligent diagnosis, artificial intelligence methods have become a research hotspot of radiography processing technology in recent years. The low definition of knee magnetic resonance image texture seriously affects the diagnosis of knee osteoarthritis. This paper presents a super-resolution reconstruction method to address this problem. MethodsIn this paper, we propose an efficient medical image super-resolution (EMISR) method, in which we mainly adopted three hidden layers of super-resolution convolution neural network (SRCNN) and a sub-pixel convolution layer of efficient sub-pixel convolution neural network (ESPCN). The addition of the efficient sub-pixel convolutional layer in the hidden layer and the small network replacement consisting of concatenated convolutions to address low-resolution images but not high-resolution images are important. The EMISR method also uses cascaded small convolution kernels to improve reconstruction speed and deepen the convolution neural network to improve reconstruction quality. ResultsThe proposed method is tested in the public dataset IDI, and the reconstruction quality of the algorithm is higher than that of the sparse coding-based network (SCN) method, the SRCNN method, and the ESPCN method (+ 2.306 dB, + 2.540 dB, + 1.089 dB improved); moreover, the reconstruction speed is faster than its counterparts (+ 4.272 s, + 1.967 s, and + 0.073 s improved). ConclusionThe experimental results show that our EMISR framework has improved performance and greatly reduces the number of parameters and training time. Furthermore, the reconstructed image presents more details, and the edges are more complete. Therefore, the EMISR technique provides a more powerful medical analysis in knee osteoarthritis examinations.

Scalable Kernelization for Maximum Independent Sets

The most efficient algorithms for finding maximum independent sets in both theory and practice use reduction rules to obtain a much smaller problem instance called a kernel . The kernel can then be solved quickly using exact or heuristic algorithms—or by repeatedly kernelizing recursively in the branch-and-reduce paradigm. Current algorithms are either slow but produce a small kernel or fast and give a large kernel. Yet it is of critical importance for these algorithms that kernelization is fast and returns a small kernel. We attempt to accomplish both of these goals simultaneously by giving an efficient parallel kernelization algorithm based on graph partitioning and parallel bipartite maximum matching. We combine our parallelization techniques with two techniques to accelerate kernelization further: dependency checking that prunes reductions that cannot be applied, and reduction tracking that allows us to stop kernelization when reductions become less fruitful. Our algorithm produces kernels that are orders of magnitude smaller than the fastest kernelization methods while having a similar execution time. Furthermore, our algorithm is able to compute kernels with size comparable to the smallest known kernels but up to two orders of magnitude faster than possible previously. Finally, we show that our kernelization algorithm can be used to accelerate existing state-of-the-art heuristic algorithms, allowing us to find larger independent sets faster on large real-world networks and synthetic instances.

Economic Analysis of Small Scale Palm Kernel Oil Producers In Ogbomosho Area of Oyo State, Nigeria

Economic Analysis of Small Scale Palm Kernel Oil Producers In Ogbomosho Area of Oyo State, Nigeria

Effects of wheat kernel size on hardness and various quality characteristics

Kernel size and hardness are important for millers and they should be controlled prior to milling. The kernel hardness also determines end-use quality of wheat. The objective of this study was to determine the effect of kernel size on wheat kernel characteristics, especially kernel hardness, milling yield and various flour quality properties. Three bread wheat varieties (Bayraktar 2000, Demir 2000 and Ikizce 96) were used after separating them into three kernel size groups. As expected, 1000 kernel weight significantly (P<0.05) decreased with decreasing kernel size. Results of three hardness tests (single kernel characterisation system, pearling index and particle size index) were parallel and hardness of the samples significantly increased (P<0.05) as the kernel size decreased. Flour yields were generally lower in the smaller kernels. However, the decrease in flour yield was not significant for cv Bayraktar 2000. In two cultivars protein content and wet gluten had a significant (P<0.05) decreasing tendency with decreasing kernel size while those of the third cultivar were not affected by kernel size. Sedimentation values of the samples were not affected by kernel size. Overall, small kernels did not seem to influence flour quality while they generally had a detrimental effect on milling yield.

Hyper-reflective foci segmentation in SD-OCT retinal images with diabetic retinopathy using deep convolutional neural networks.

The purpose of this study was to automatically and accurately segment hyper-reflective foci (HRF) in spectral domain optical coherence tomography (SD-OCT) images with diabetic retinopathy (DR) using deep convolutional neural networks. An automatic HRF segmentation model for SD-OCT images based on deep networks was constructed. The model segmented small lesions through pixel-wise predictions based on small image patches. We used an approach for discriminative features extraction for small patches by introducing small kernels and strides in convolutional and pooling layers, which was applied on the state-of-the-art deep classification networks (GoogLeNet and ResNet). The features extracted by the adapted deep networks were fed into a softmax layer to produce the probabilities of HRF. We trained different models on a dataset with 16 HRF eyes by using different sizes of patches, and then, we fused these models to generate optimal results. Experimental results on 18 eyes demonstrated that our method is effective for the HRF segmentation. The dice similarity coefficient (DSC) for the foci area in B-scan, projection images, and foci amount in B-scan images reaches 67.81%, 74.09%, and 72.45%, respectively. The proposed segmentation model can accurately segment HRF in SD-OCT images with DR and outperforms traditional methods. Our model may provide reliable segmentations for small lesions in SD-OCT images and may be helpful in the clinical diagnosis of diseases.

Estimating the Major Cluster by Mean-Shift with Updating Kernel

The mean-shift method is a convenient mode-seeking method. Using a principle of the sample mean over an analysis window, or kernel, in a data space where samples are distributed with bias toward the densest direction of sample from the kernel center, the mean-shift method is an attempt to seek the densest point of samples, or the sample mode, iteratively. A smaller kernel leads to convergence to a local mode that appears because of statistical fluctuation. A larger kernel leads to estimation of a biased mode affected by other clusters, abnormal values, or outliers if they exist other than in the major cluster. Therefore, optimal selection of the kernel size, which is designated as the bandwidth in many reports of the literature, represents an important problem. As described herein, assuming that the major cluster follows a Gaussian probability density distribution, and, assuming that the outliers do not affect the sample mode of the major cluster, and, by adopting a Gaussian kernel, we propose a new mean-shift by which both the mean vector and covariance matrix of the major cluster are estimated in each iteration. Subsequently, the kernel size and shape are updated adaptively. Numerical experiments indicate that the mean vector, covariance matrix, and the number of samples of the major cluster can be estimated stably. Because the kernel shape can be adjusted not only to an isotropic shape but also to an anisotropic shape according to the sample distribution, the proposed method has higher estimation precision than the general mean-shift.

A Mitochondrial Transcription Termination Factor, ZmSmk3, Is Required for nad1 Intron4 and nad4 Intron1 Splicing and Kernel Development in Maize.

The expression systems of the mitochondrial genes are derived from their bacterial ancestors, but have evolved many new features in their eukaryotic hosts. Mitochondrial RNA splicing is a complex process regulated by families of nucleus-encoded RNA-binding proteins, few of which have been characterized in maize (Zea mays L.). Here, we identified the Zea mays small kernel 3 (Zmsmk3) candidate gene, which encodes a mitochondrial transcription termination factor (mTERF) containing two mTERF motifs, which is conserved in monocotyledon; and the target introns were also quite conserved during evolution between monocotyledons and dicotyledons. The mutations of Zmsmk3 led to arrested embryo and endosperm development, resulting in small kernels. A transcriptome of 12 days after pollination endosperm analysis revealed that the starch biosynthetic pathway and the zein gene family were down-regulated in the Zmsmk3 mutant kernels. ZmSMK3 is localized in mitochondria. The reduced expression of ZmSmk3 in the mutant resulted in the splicing deficiency of mitochondrial nad4 intron1 and nad1 intron4, causing a reduction in complex I assembly and activity, impairing mitochondria structure and activating the alternative respiratory pathway. So, the results suggest that ZmSMK3 is required for the splicing of nad4 intron 1 and nad1 intron 4, complex I assembly and kernel development in maize.

ZmSMK9, a pentatricopeptide repeat protein, is involved in the cis-splicing of nad5, kernel development and plant architecture in maize

Maize kernel size and weight are essential contributors to its yield. So the identification of the genes controlling kernel size and weight can give us a chance to gain the yield. Here, we identified a small kernel mutant, Zea mays small kernel 9 (Zmsmk9), in maize. Cytological observation showed that the development of the endosperm and embryo was delayed in Zmsmk9 mutants at the early stages, resulting in a small kernel phenotype. Interestingly, despite substantial variation in kernel size, the germination of Zmsmk9 seeds was comparable to that of WT, and could develop into normal plants with upright leaf architecture. We cloned Zmsmk9 via map-based cloning. ZmSMK9 encodes a P-type pentatricopeptide repeat protein that targets to mitochondria, and is involved in RNA splicing in mitochondrial NADH dehydrogenase5 (nad5) intron-1 and intron-4. Consistent with the delayed development phenotype, transcriptome analysis of 12-DAP endosperm showed that starch and zeins biosynthesis related genes were dramatically down regulated in Zmsmk9, while cell cycle and cell growth related genes were dramatically increased. As a result, ZmSMK9 is a novel gene required for the splicing of nad5 intron-1 and intron-4, kernel development, and plant architecture in maize.

Compressed sensing MRI via a multi-scale dilated residual convolution network

Magnetic resonance imaging (MRI) reconstruction is an active inverse problem which can be addressed by conventional compressed sensing (CS) MRI algorithms that exploit the sparse nature of MRI in an iterative optimization-based manner. However, two main drawbacks of iterative optimization-based CSMRI methods are time-consuming and are limited in model capacity. Meanwhile, one main challenge for recent deep learning-based CSMRI is the trade-off between model performance and network size. To address the above issues, we develop a new multi-scale dilated network for MRI reconstruction with high speed and outstanding performance. Comparing to convolutional kernels with same receptive fields, dilated convolutions reduce network parameters with smaller kernels and expand receptive fields of kernels to obtain almost same information. To maintain the abundance of features, we present global and local residual learnings to extract more image edges and details. Then we utilize concatenation layers to fuse multi-scale features and residual learnings for better reconstruction. Compared with several non-deep and deep learning CSMRI algorithms, the proposed method yields better reconstruction accuracy and noticeable visual improvements. In addition, we perform the noisy setting to verify the model stability, and then extend the proposed model on a MRI super-resolution task.

Brain Tumor Segmentation Using Convolutional Neural Networks in MRI Images.

In medical image processing, Brain tumor segmentation plays an important role. Early detection of these tumors is highly required to give Treatment of patients. The patient's life chances are improved by the early detection of it. The process of diagnosing the brain tumoursby the physicians is normally carried out using a manual way of segmentation. It is time consuming and a difficult one. To solve these problems, Enhanced Convolutional Neural Networks (ECNN) is proposed with loss function optimization by BAT algorithm for automatic segmentation method. The primary aim is to present optimization based MRIs image segmentation. Small kernels allow the design in a deep architecture. It has a positive consequence with respect to overfitting provided the lesser weights are assigned to the network. Skull stripping and image enhancement algorithms are used for pre-processing. The experimental result shows the better performance while comparing with the existing methods. The compared parameters are precision, recall and accuracy. In future, different selecting schemes can be adopted to improve the accuracy.

Understanding the diesel-like spray characteristics applying a flamelet-based combustion model and detailed large eddy simulations

This investigation analyses the structure of spray A from engine combustion network (ECN), which is representative of diesel-like sprays, by means of large eddy simulations and an unsteady flamelet progress variable combustion model. A very good agreement between modelled and experimental measurements is obtained for the inert spray that supports further analysis. A parametric variation in oxygen concentration is carried out in order to describe the structure of the flame and how it is modified when mixture reactivity is changed. The most relevant trends for the flame metrics, ignition delay and lift-off length are well-captured by the simulations corroborating the suitability of the model for this type of configuration. Results show that the morphology of the flame is strongly affected by the boundary conditions in terms of the reactive scalar spatial fields and Z–T maps. The filtered instantaneous fields provided by the simulations allow investigation of the structure of the flame at the lift-off length, whose positioning shows low fluctuations, and how it is affected by turbulence. It is evidenced that small ignition kernels appear upstream and detached from the flame that eventually merge with its base in agreement with experimental observations, leading to state that auto-ignition plays a key role as one of the flame stabilization mechanisms of the flame.

A novel adaptive and fast deep convolutional neural network for bearing fault diagnosis under different working conditions

Deep learning method is gradually applied in the field of mechanical equipment fault diagnosis because it can learn complex and useful features automatically from the vibration signals. Among the many intelligent diagnostic models, convolutional neural network has been gradually applied to intelligent fault diagnosis of bearings due to its advantages of local connection and weight sharing. However, there are still some drawbacks. (1) The training process of convolutional neural network is slow and unstable. It has more training parameters. (2) It cannot perform well under different working conditions, such as noisy environment and different workloads. In this paper, a novel model named adaptive and fast convolutional neural network with wide receptive field is presented to overcome the aforementioned deficiencies. The prime innovations include the following. First, a deep convolutional neural network architecture is constructed using the scaled exponential linear unit activation function and global average pooling. The model has fewer training parameters and can converge rapidly and stably. Second, the model has a wide receptive field with two medium and three small length convolutional kernels. It also has high diagnostic accuracy and robustness when the environment is noisy and workloads are changed compared with other models. Furthermore, to demonstrate how the wide receptive field convolutional neural network model works, the reasons for high model performance are analyzed and the learned features are also visualized. Finally, the wide receptive field convolutional neural network model is verified by the vibration dataset collected in the background of high noise, and the results indicate that it has high diagnostic performance.