Local Density Maxima Research Articles

Maximum Local Density-Driven Non-overlapping Radial Basis Function Support Kernel Neural Network

The learning and optimization of kernels in the radial basis function neural network (RBFNN) are crucial. However, in existing methods, there are issues of overfitting when learning kernel parameters. The learned kernels are also sensitive to outliers. This paper proposes a general kernel learning strategy for RBFNN called non-overlapping maximum local density support kernel learning (MLD-SKL), which contains two modules, the non-overlapping maximum local density (MLD) kernel learning module and support kernel learning (SKL) module. In the MLD kernel learning stage, the candidate set of kernels is incrementally determined based on the local density of samples. Meanwhile, it is required that the coverage ranges of kernels from different classes do not overlap with each other. This module is effective in reducing the impact of outliers. In the SKL stage, kernel importance indicator is defined to measure the importance of kernels. The learned support kernels are utilized to construct a maximum local density-driven non-overlapping radial basis function support kernel neural network (MLD-RBFSKNN). The RBFNN constructed through MLD-SKL exhibits a more compact structure. The experiments demonstrate that the proposed MLD-RBFSKNN improves accuracy in recognition task. Furthermore, while achieving superior recognition performance, the final constructed network also has the minimum number of kernels.

Frequency Dependence of the Parameters of the Inductive RF Discharge Located in the Low-Value Magnetic Field

In this work, we carried out studies of the properties of an inductive RF discharge placed in a magnetic field with an induction of less than 70 G at frequencies of 2, 4 and 13.56 MHz. Experiments have shown that when operating at frequencies of 2 and 4 MHz at low powers of the RF generator, the range of existence of the discharge is limited by large magnetic fields. The efficiency of RF power input η non-monotonically depends on the magnitude of the magnetic field. The position of the main maximum η shifts to the region of higher B with increasing frequency, power of the RF generator and argon pressure, and at the same time the maximum broadens. An increase in frequency, power and argon pressure is accompanied by an increase in the absolute values of η. When operating at a frequency of 4 MHz, in addition to the main maximum η, a local maximum appears in the region B 35–70 G. With increasing pressure, a shift in the position of the local maximum and its smoothing is observed. Comparison of experimental data with calculated data allows us to conclude that the local maximum of plasma density observed at weak magnetic fields is associated with resonant excitation of waves in the plasma source. At a frequency of 2 MHz, the excited wave is close to a transverse helicon, and at a frequency of 13.56 MHz, its properties approach the Trivelpiece–Gold wave.

Substantial breakdown of the hydrogen-bonding network, local density inhomogeneities and fluid-liquid structural transitions in supercritical octanol-1: A molecular dynamics investigation.

Molecular dynamics simulations have been employed to explore the hydrogen-bonding structure and dynamics in supercritical octanol-1 at a near-critical temperature and up to high densities and pressures. A substantial breakdown of the hydrogen-bonding network when going from ambient-liquid to supercritical conditions is revealed. The fraction of the non-hydrogen bonded molecules significantly increases in supercritical octanol-1, and a substantial decrease in the intermittent hydrogen-bond lifetime is observed. This behavior is also reflected on the maximum local density augmentation, which is comparable to the values obtained for non-polar and non-hydrogen bonded fluids. The existence of a structural transition from an inhomogeneous fluid phase to a soft-liquid one at densities higher than 2.0 ρc is also revealed. At higher densities, a significant change in the reorientational relaxation process is observed, reflected on the significant increase in the ratio of the Legendre reorientational times τ1R/τ2R. The latter becomes much higher than the value predicted by the Debye model of diffusive reorientation and the corresponding ratio for ambient liquid octanol-1. The non-polar tail of octanol-1 under supercritical conditions reorients more slowly in comparison with the polar tail. Interestingly, the opposite behavior is observed for the ambient liquid, further verifying the strong effect of the breakdown of the hydrogen bonding network on the properties of supercritical octanol-1. In accordance with the above-mentioned findings, the static dielectric constant of supercritical octanol-1 is very low even at high densities and pressures, comparable to the values obtained for non-polar and non-hydrogen bonded fluids.

Effect of Shock-Wave-Mediated Collapse on Nanobubble Characteristics.

To enhance the comprehension of the cavitation mechanism and explore its practical use in industrial production, this study developed models involving oxygen, varying bubble radii, and bubble quantities. This study uses molecular dynamics simulations coupled with the momentum mirror method to examine the collapse characteristics of bubbles during ultrasonic cavitation. The investigation uncovers patterns in the fluctuation of the maximum local density of water molecules, released pressure, and temperature. The findings demonstrate that, when oxygen-containing bubbles collapse at identical radii, the local density is notably higher and diminishes more rapidly. Moreover, the changes in the shape exhibit greater regularity. During the bubble collapse, a depression forms on the bubble's surface, coinciding with a notable surge in local density around the depression. As bubble radii and quantities increase, so does the local density along with a concurrent rise in the maximum pressure. Intriguingly, the model demonstrates the lowest pressure at Z = 35 Å accompanied by the emergence of a small crescent-shaped region with a reduced density. Throughout the pressure ascension phase, the rate of the maximum pressure change escalates with an increase in the number of bubbles. Conversely, during the pressure descent phase, the rate of the maximum pressure change diminishes with a growing number of bubbles. However, it is important to note that the maximum pressure does not exhibit a direct correlation with the number of bubbles. Ultimately, this study provides valuable technical guidance and a theoretical foundation for the integration of ultrasonic cavitation in industrial production processes.

Interaction phenomena of electrically coupled cells under local reactant starvation in automotive PEMFC stacks

The local current density distribution of polymer electrolyte membrane fuel cells (PEMFCs) can be distorted by various error states. Differences in current density distributions (CDDs) of adjacent cells in a stack are equilibrized by in-plane currents within the sandwiched bipolar plates. Degradation stressors such as detrimental differences in local cell voltage and current density maxima can thus be generated. A novel method was therefore developed to intentionally manipulate CDD profiles by integrating local artificial starvation into only one fuel cell in an assembly. This technique is applied to automotive-sized PEMFCs single cells as well as in 20 cell short-stack to analyze such voltage and current redistribution phenomena. A drastic distortion of local cell voltage is only observed for stacks, which is explained by a supplementary simulation. The local voltage distribution of an electrically coupled fuel cell is therefore calculated by combining CDD measurements with a spatially resolved polarization curve model. The capabilities and limits of a multipoint cell voltage monitoring measurement device are discussed on this basis. The inspected correlation between these two independent online measurement techniques allows to localize such error states with considerable accuracy during operation of automotive sized PEMFC stacks.

Strong Scattering Effects in the Emission of Soft Gamma-Ray Bursts

When a light beam enters a scattering-dominated medium, the radiation is isotropized. Part of the radiation goes backwards, leading to non-monotonicity in the radiation energy density profile inside this medium. There arises a local maximum at which the energy density at a scattering albedo 1 is severalfold greater than that without scattering at the same extinction. This effect is studied numerically in one-dimensional and two-dimensional simulations. It is demonstrated that a local maximum of the radiation energy density arises in the medium, whose value depends on the optical depth of the region. This effect can manifest itself, for example, when the radiation from a gamma-ray burst (GRB) enters heated regions in the interstellar medium. The presence of scattering in the GRB radiation generation region, near the front of strong shocks, affects the radiation pattern. The structure of such shocks is remarkable for the presence of a preshock preheating tail. Strong scattering in this region leads to the escape of a significant fraction of the radiation sideways and backwards in the shock reference frame, forming additional tails in the angular distribution of GRB radiation after the relativistic transformation to the laboratory frame. This effect is also studied numerically in three-dimensional simulations.

Identification of Areas of Anomalous Tremor of the Earth’s Surface on the Japanese Islands According to GPS Data

Statistical properties of Earth surface tremors measured by means of GPS were investigated. This article considers measurements of the Earth’s surface displacements in three orthogonal directions relayed by a network of GPS sensors with about 1200 points distributed across Japan in 2009–2021. Next, the following characteristics of the tremors were considered: the entropy of the distribution of squared orthogonal wavelet coefficients, the entropy of the distribution of power spectrum values, and the spectral index. The anomalous regions of maxima of probability densities of the distribution of extreme values of the tremor statistics were determined: entropy minima and spectral index maxima. The average density maps of the distribution of extreme value tremor statistics were found to be highly correlated with one another. This made it possible to consider a weighted average density map and identify five anomalous regions in the center and south of Japan. A trajectory of visiting anomalous regions by a sequence of points realizing local maxima of the average probability density was obtained, for which seasonal periodicity was set. Estimates of changes in the average and maximum values of the correlation coefficients of tremor properties in an auxiliary network of 16 reference points in a semi-annual time window were obtained.

Geometric Predictors of Knotted and Linked Arcs.

Inspired by how certain proteins “sense” knots and entanglements in DNA molecules, here, we ask if local geometric features that may be used as a readout of the underlying topology of generic polymers exist. We perform molecular simulations of knotted and linked semiflexible polymers and study four geometric measures to predict topological entanglements: local curvature, local density, local 1D writhe, and nonlocal 3D writhe. We discover that local curvature is a poor predictor of entanglements. In contrast, segments with maximum local density or writhe correlate as much as 90% of the time with the shortest knotted and linked arcs. We find that this accuracy is preserved across different knot types and also under significant spherical confinement, which is known to delocalize essential crossings in knotted polymers. We further discover that nonlocal 3D writhe is the best geometric readout of the knot location. Finally, we discuss how these geometric features may be used to computationally analyze entanglements in generic polymer melts and gels.

Comparing results of real-scale time MHD modeling with observational data for first flare M 1.9 in AR 10365

Abstract As shown in the first results of MHD simulations in the real scale of time, above the active region (AR) 10365, during the first flare M 1.9 (05/26/2003 05:34) at a height of 16–18 mm (lower corona), a singular line of magnetic field appears. The local maximum of the current density is situated on this singular line. The magnetic field in the vicinity of this singular line is the superposition of an X-type magnetic configuration and a divergent magnetic field. The accumulation of magnetic energy for solar flare with current sheet creation takes place near this singular line due to magnetic field deformation by disturbances in the X-type configuration in spite of the presence of overlaid diverging magnetic configuration. The magnetic configuration is so complicated that the singular line can be found only by using specially developed graphical system of search. The position of singular line coincides with position of source of flare radio emission at the frequency 17 GHz above AR 10365 measured by Nobeyama Radioheliograph (NoRH). Also, MHD simulation shows appearance of the singular line, in the vicinity of which X-type configuration dominates. However, apparently due to small disturbance, propagating from the photosphere, sufficient magnetic energy was not accumulated in this configuration, so the NoRH does not show the flare source of emission at the frequency 17 GHz in the place, where this singular line is situated.

Multi-density peaks clustering superpixel

A superpixel segmentation algorithm called multi-density peaks clustering (MDPC) is proposed. By selecting a sufficient number of local density maximum pixels from the image as cluster centers to depict the image texture, the boundary of the object can be captured very accurately. The algorithm framework of MDPC is divided into three steps. First, the local density of pixel is defined, and the local density maximum pixels are calculated. Then, the local density maximum pixels are used as cluster centers, and the global optimal search, which is based on the path-to-point idea, is used to complete the clustering of the remaining non-cluster center pixels to realize the initial segmentation. Finally, superpixels are obtained by merging the initial segments according to the size of the segments and the distance between adjacent segments. In quantitative comparisons, MDPC is compared with 13 state-of-the-art superpixel segmentation algorithms in three image segmentation datasets. The experimental results show that MDPC achieves better performance in terms of boundary recall, boundary precision, achievable segmentation accuracy, undersegmentation error, and explained variation. And the qualitative comparisons show that the proposed algorithm has obvious advantages over other superpixel segmentation algorithms in image detail description and boundary adherence. Finally, the practicability and stability of MDPC are further demonstrated by the application of image segmentation. The source code of MDPC will be available at https://github.com/zhaojianaaa.

Electron power absorption in micro atmospheric pressure plasma jets driven by tailored voltage waveforms in He/N2

In atmospheric pressure capacitively-coupled microplasma jets, voltage waveform tailoring (VWT) was demonstrated to provide ultimate control of the electron energy distribution function (EEDF), which allows us to enhance and adjust the generation of selected neutral species by controlling the electron power absorption dynamics. However, at the fundamental level, the physical origin of these effects of VWT remained unclear. Therefore, in this work, the electron power absorption dynamics is investigated in a He/N2 jet with a nitrogen concentration of 0.05% driven by a valleys voltage waveform at a base frequency of 13.56 MHz for different numbers of harmonics, using a self-consistent particle-in-cell simulation coupled with a spatio-temporally resolved analysis of the electron power absorption based on moments of the Boltzmann equation. Due to the local nature of the transport at atmospheric pressure, ohmic power absorption is dominant. Increasing the number of harmonics, due to the peculiar shape of the excitation waveform, the sheath collapse at the grounded electrode is shortened relative to the one at the powered electrode. As a consequence, and in order to ensure flux compensation of electrons and positive ions at this electrode, a high current is driven through the discharge at the time of this short sheath collapse. This current is primarily driven by a high ohmic electric field. Close to the grounded electrode, where the electron density is low and the electric field is therefore high, electrons are accelerated to high energies and strong ionization, as well as the formation of a local electron density maximum, are observed. This electron density maximum leads to a local ambipolar electric field that acts as an electric field reversal and accelerates electrons to even higher energies. These effects are understood in detail to fundamentally explain the unique potential of VWT for EEDF control in such plasmas.

Urban hotspots detection of taxi stops with local maximum density

When getting on and off the taxis, people prefer to choose around landmarks, such as a shopping mall, the gate of a residential unit, or a road intersection. This preference leads to multiple local peaks in spatial density, existing in both highly popular and less popular regions. These multiple local peaks provide a natural form to describe small-scale hotspots where the pick-up and drop-off events actually happen. However, they have not been fully studied. In this paper, we propose the local maximum density (LMD) approach to identify the hotspots as a small area around a local maximum density of incidents. It is evaluated using a 6-month taxi dataset of Wuhan City, where 90 m × 90 m square is recommended as the size of a hotspot by LMD. Results show that LMD not only identifies multiple local hotspots in highly popular regions, but also detects potential hotspots in less popular regions. Moreover, a non-uniform spatial pattern is found between pick-up and drop-off local hotspots. Comparing with pick-up behaviors, drop-off behaviors diffuse more in less popular hotspots, but also concentrate more on highly popular hotspots. The driving factors of this phenomenon are further explored by analyzing the match mode of pick-up and drop-off local hotspots with different popularity.

Influence of an external circuit on the plasma parameters in the channel of the radio-frequency accelerator with a closed electron drift

The axial distribution of the plasma potential, concentration and temperature of electrons in an RF capacitive plasma source with the geometry of an accelerator with a closed electron drift is experimentally investigated in this work. Two cases of the external electrical discharge circuit are considered. In the first case, the electrodes were closed by direct current, in the second, they were opened. It is shown that direct current closure of the electrodes leads to a significant increase in the plasma potential and electron concentration. In a number of cases, local maxima of temperature and plasma density are observed near the electrodes, which can be associated with the occurrence of azimuthal electron drift in crossed electric and magnetic fields.

Influence of an external circuit on the plasma parameters in the channel of the radio-frequency accelerator with a closed electron drift

The axial distribution of the plasma potential, concentration and temperature of electrons in an RF capacitive plasma source with the geometry of an accelerator with a closed electron drift is experimentally investigated in this work. Two cases of the external electrical discharge circuit are considered. In the first case, the electrodes were closed by direct current, in the second, they were opened. It is shown that direct current closure of the electrodes leads to a significant increase in the plasma potential and electron concentration. In a number of cases, local maxima of temperature and plasma density are observed near the electrodes, which can be associated with the occurrence of azimuthal electron drift in crossed electric and magnetic fields.

An Improved Integrated Clustering Learning Strategy Based on Three-Stage Affinity Propagation Algorithm with Density Peak Optimization Theory

To better reflect the precise clustering results of the data samples with different shapes and densities for affinity propagation clustering algorithm (AP), an improved integrated clustering learning strategy based on three-stage affinity propagation algorithm with density peak optimization theory (DPKT-AP) was proposed in this paper. DPKT-AP combined the ideology of integrated clustering with the AP algorithm, by introducing the density peak theory and k-means algorithm to carry on the three-stage clustering process. In the first stage, the clustering center point was selected by density peak clustering. Because the clustering center was surrounded by the nearest neighbor point with lower local density and had a relatively large distance from other points with higher density, it could help the k-means algorithm in the second stage avoiding the local optimal situation. In the second stage, the k-means algorithm was used to cluster the data samples to form several relatively small spherical subgroups, and each of subgroups had a local density maximum point, which is called the center point of the subgroup. In the third stage, DPKT-AP used the AP algorithm to merge and cluster the spherical subgroups. Experiments on UCI data sets and synthetic data sets showed that DPKT-AP improved the clustering performance and accuracy for the algorithm.

Effects of driving frequency on plasma density in Ar and H2/Ar capacitively coupled plasmas at Torr-order pressure

Effects of driving frequency on the plasma density in capacitively coupled plasmas (CCPs) have been investigated at Torr-order pressures which are commonly used in film deposition processes (1 Torr = 133.322 Pa). Using a particle-in-cell, Monte Carlo collision model, frequency trends of the electron density have been simulated in Ar and H2/Ar CCPs from 300 kHz to 27 MHz. The results show a local maximum of the electron density at 450 kHz and a local minimum around 4 MHz in both the Ar and H2/Ar plasmas. In particular, in the H2/Ar plasma, 450 kHz is more effective to produce higher density plasma than 13.56 MHz at Torr-order pressure although it is generally believed that higher density plasma can be obtained as the driving frequency increases. The ion density starts to oscillate between electrodes around 4 MHz since it is close to the ion plasma frequency. Thus, the plasma bulk collapses resulting in the low plasma density. On the other hand, at 450 kHz, in addition to the gamma process, the re-formed plasma bulk contributes to high density plasma.

Standing autoresonant plasma waves

The formation and control of strongly nonlinear standing plasma waves (SPWs) from a trivial equilibrium by a chirped frequency drive are discussed. If the drive amplitude exceeds a threshold, after passage through the linear resonance in this system, the excited wave preserves the phase locking with the drive, yielding a controlled growth of the wave amplitude. We illustrate these autoresonant waves via Vlasov–Poisson simulations, showing the formation of sharply peaked excitations with local electron density maxima significantly exceeding the unperturbed plasma density. The Whitham averaged variational approach applied to a simplified water bag model yields the weakly nonlinear evolution of the autoresonant SPWs and the autoresonance threshold. If the chirped driving frequency approaches some constant level, the driven SPW saturates at a target amplitude, avoiding the kinetic wave breaking.

Differential diffusion effects on density-driven instability of reactive flows in porous media

Based on the lattice Boltzmann (LB) method, pore-scale simulations are performed to investigate the differential diffusion effects on the density-driven instability (DI) with chemical reaction A + B → C in porous media. A partially miscible stratification is considered, and thus only solutes from the top fluid can diffuse down into the host fluid in pore spaces. Tests with different values of the Rayleigh number Ra r and the diffusion coefficient D r of species r ( r = A , B , C ) are considered. The results demonstrate eight distinct scenarios of DI, and four of them are not observed in equal diffusivity simulations. Two differential diffusion effects, namely, the double-diffusive (DD) and the diffusive-layer convection (DLC) mechanisms, can act upon the gravity field and give rise to new fingering phenomena. The DD mechanism comes into play and results in a local minimum density layer if Ra B / Ra C is small and D B > D C ; and DLC becomes significant and brings in a local maximum density layer if Ra B / Ra C is large and D B < D C . On one hand, when fluid density increases with dissolved A , the DD-induced minimum can act as an inhibiting barrier to suppress fingering propagation, although it can be eventually penetrated by fingering tips; and the DLC-induced maximum can introduce the second DI below the first one. On the other hand, when the dissolution of A contributes to decreasing fluid density, both the DD-induced minimum and the DLC-induced maximum can help trigger the development of DI. Finally, quantitative results are provided to indicate that fingering propagates into the host fluid more deeply with larger D B / D C , and the dissolution of A decreases with the increasing difference between D B and D C .

De Novo Protein Structure Modeling Tool MAINMAST Enhanced for Multiple Chain Complexes and Bound Ligands

The significant progress of the cryo-EM poses a pressing need for software for structural interpretation of EM maps. Particularly, protein structure modeling tools are needed for maps determined around 4 Å resolution, where modeling protein structure in EM map are still challenging problems. We have developed a de novo modeling tool named MAINMAST (MAINchain Model trAcing from Spanning Tree, http://kiharalab.org/mainmast/) for EM maps for this resolution range (Nature Communications, 2018). MAINMAST builds main-chain traces of a protein in an EM map from a tree graph structure constructed by connecting local maximum density points (LDPs)in the EM map. Here, we report the new version of MAINMAST that can handle (1) multiple protein chains with symmetry and (2) proteins with bound ligands in EM maps. For modeling of multiple protein chains, the new MAINMAST automatically identifies pairs of LDPs that locate in symmetric positions. Then, the identified LDP pairs are recorded as a tabu-pair list, which should not be included in a single chain. With the computed restrains in the tabu-pair list, MAINMAST constructs minimum spanning trees, which satisfy the all restrains and will produce symmetric chain structures. Once the initial trees are generated, the trees are optimized, and then the target sequence is mapped. For modeling of proteins with bound ligands, MAINMAST analyzes the tree-graph data and constructs main-chain model, and then detects the ligand-binding sites. Once MAINMAST identify the potential position of the bound ligand, MAINMAST segments the density map around the binding site, and models the ligand structure. We tested the new protocol on 24-mer complexes, apo-ferritin (emd-20026, emd-20027 and emd-20028) and a dimer complex, alcohol dehydrogenase with NAD molecules (emd-0406). MAINMAST built accurate protein structure models and NAD ligand models.

A new internal index based on density core for clustering validation

Clustering validation which is applied to the evaluation of clustering results has been recognized as one of the vital issues for clustering application. Density core, a set of connected maximum local density peaks, can approximately represent the structure of a cluster without being affected by noises. Therefore, a density-core-based clustering validation index (DCVI) with minimum spanning tree (MST) is proposed to solve the problem that noises and arbitrary shapes have great influence on the performance of widely used internal validation measures. As for data sets containing arbitrarily shaped clusters with noises, DCVI can effectively obtain the optimal number of clusters. Moreover, experimental results of both synthetic and real data sets indicate that DCVI outperforms most existing measures.