Single-level Structure Research Articles

Hierarchical multi-level dynamic hyperparameter deformable image registration with convolutional neural network

Objective. To enable the registration network to be trained only once, achieving fast regularization hyperparameter selection during the inference phase, and to improve registration accuracy and deformation field regularity. Approach. Hyperparameter tuning is an essential process for deep learning deformable image registration (DLDIR). Most DLDIR methods usually perform a large number of independent experiments to select the appropriate regularization hyperparameters, which are time-consuming and resource-consuming. To address this issue, we propose a novel dynamic hyperparameter block, which comprises a distributed mapping network, dynamic convolution, attention feature extraction layer, and instance normalization layer. The dynamic hyperparameter block encodes the input feature vectors and regularization hyperparameters into learnable feature variables and dynamic convolution parameters which changes the feature statistics of the high-dimensional features layer feature variables, respectively. In addition, the proposed method replaced the single-level structure residual blocks in LapIRN with a hierarchical multi-level architecture for the dynamic hyperparameter block in order to improve registration performance. Main results. On the OASIS dataset, the proposed method reduced the percentage of |Jϕ|⩽0 by 28.01 % , 9.78 % and improved Dice similarity coefficient by 1.17 % , 1.17 % , compared with LapIRN and CIR, respectively. On the DIR-Lab dataset, the proposed method reduced the percentage of |Jϕ|⩽0 by 10.00 % , 5.70 % and reduced target registration error by 10.84 % , 10.05 % , compared with LapIRN and CIR, respectively. Significance. The proposed method can fast achieve the corresponding registration deformation field for arbitrary hyperparameter value during the inference phase. Extensive experiments demonstrate that the proposed method reduces training time compared to DLDIR with fixed regularization hyperparameters while outperforming the state-of-the-art registration methods concerning registration accuracy and deformation smoothness on brain dataset OASIS and lung dataset DIR-Lab.

Compound nested lattices with programmable isotropy and elastic stiffness up to the theoretical limit

A novel class of compound structures, which consists of 2 types of unit cell geometries occupying different sites in the lattice (i.e. compound lattice) was investigated. The arrangement and volume ratio of the 2 unit cell geometries were varied, and it was found that the compound lattices can exhibit up to 4 distinct geometries – 2 from the unit cells and 2 supra-structures from the arrangement of each type of unit cell. In stiffness optimization, the material re-organization tends to emphasize the stiffest of the 4 geometries and collapse the hierarchical compound lattice into a single-level structure. In isotropy optimization, unit cells had to be arranged into supra-structures with an anisotropy profile opposite to that of their geometries. These insights led to the introduction of the compound nested lattices, which exhibited higher specific moduli than previous isotropic designs. The compound nested 1pSC:512pFCC lattice, in particular, reached 97.9 % of the Hashin-Shtrikman upper bound at relative density = 0.6, which is the closest approach to the theoretical maximum ever reported.

Hierarchical overlapping modular structure in the human cerebral cortex improves individual identification

The idea that brain networks have a hierarchical modular organization is pervasive. Increasing evidence suggests that brain modules overlap. However, little is known about the hierarchical overlapping modular structure in the brain. In this study, we developed a framework to uncover brain hierarchical overlapping modular structures based on a nested-spectral partition algorithm and an edge-centric network model. Overlap degree between brain modules is symmetrical across hemispheres, with highest overlap observed in the control and salience/ventral attention networks. Furthermore, brain edges are clustered into two groups: intrasystem and intersystem edges, to form hierarchical overlapping modules. At different levels, modules are self-similar in the degree of overlap. Additionally, the brain's hierarchical structure contains more individual identifiable information than a single-level structure, particularly in the control and salience/ventral attention networks. Our results offer pathways for future studies aimed at relating the organization of hierarchical overlapping modules to brain cognitive behavior and disorders.

Retail Electricity Pricing Strategy via an Artificial Neural Network-Based Demand Response Model of an Energy Storage System

The distribution company (DISCO) determines optimal retail prices to operate the distribution network efficiently while promoting demand response (DR) programs. In addition, an energy storage system (ESS), which improves peak load management, is widely used for price-based DR. This paper proposes an electricity retail pricing strategy that considers the optimal operation of an ESS using a machine learning algorithm. An artificial neural network (ANN) is used to develop a practical model of the DR scheduling of an ESS. This model is trained using historical data that include the electricity price and the corresponding optimal demand obtained from the building energy management system. The proposed model is replicated using mathematical equations and directly integrated into the constraints of the retail pricing optimization problem of the distribution management system. The proposed ANN-based DR model of the ESS allows the development of an optimal pricing strategy with a single-level structure while reflecting the decision-making process of both the DISCO and the building operator. The proposed ANN-based DR model is verified through case studies, which prove that the model successfully expresses the price-optimal demand function and has high practical applicability. The results of the retail pricing demonstrate that the proposed strategy can accurately determine the balancing points while reducing the peak load.

Mathematical model of optimization of preventive maintenance of transmitting active phased antenna arrangements of multifunctional radars

In the article, the authors discuss the issues of determining the reliability indicators and maintenance of transmitting active phased array antennas (AFAR) of multifunctional radars. The analysis of the existing models of reliability and maintenance of AFAR radar. It is shown that the most interesting models of transmitting AFAR reliability are the reliability model with a two-level block diagram of the AFAR and the reliability model according to the criterion of permissible reduction of the maximum radar range. Among the AFAR maintenance models, the most interesting ones were identified: the AFAR optimal preventive maintenance (ON) model with a single-level structure and the AFAR maintenance model by technical condition. Using the results of the analysis, models of reliability and optimal preventive maintenance of a multi-functional radar transmitting AFAR are constructed. The analytical expressions of the reliability indicators of a two-level AFAR are presented: the average operating time to failure and the probabilities of failure-free operation of the antenna transmitting sublattice and the transmitting AFAR as a whole. Analytical expressions are constructed for the criteria for the effectiveness of optimal software (coefficient of technical use and mathematical expectation of specific operating costs) and the criteria for gaining from software optimization. The work considers an illustrative example of calculating the reliability and optimal software of a two-level AFAR, including 6400 microwave emitting channels, arranged in 100 sublattices with 64 channels each. The mean operating time to failure and the optimal periodicity of the AFAR software were determined for three versions of the reliability indicators of the emitting channels. The graphs of the probability of failure-free operation, the coefficient of technical use and the mathematical expectation of the specific operational costs of the AFAR were constructed and analyzed. The values of the optimal periodicity and the gain from optimization are obtained. It is shown that the results of optimal software should be used at the initial stage of operation of radars with AFAR — at the stage of warranty service. At further stages of operation - the stages of post-warranty service, it is recommended to use the AFAR maintenance method according to the technical condition. The models proposed by the authors can be useful to engineers and scientists developing radars with AFAR and employees of operating organizations.

High-temperature oxidation and plasma torch testing of MoSi2–HfB2–MoB ceramics with single-level and two-level structure

Influence of the composition and structure of the heterophase ceramics in system MoSi2–HfB2–MoB on the thermal conductivity, kinetics and mechanisms of oxidation at 1200 °C and 1650 °C, including the plasma torch testing at 2000 °C, was investigated. Dense ceramics with heterogeneous single-level structure (SLS) and two-level structure (TLS) were obtained by the hot pressing of powders produced by self-propagating high-temperature synthesis (SHS) in combustion mode. TLS ceramics have a lower thermal conductivity as compared to the SLS ceramics with similar elemental composition. Addition of hafnium diboride leads to the increase of mass change during the oxidation due to the formation of HfO2 and HfSiO4, which are denser than SiO2. In the case of the SLS ceramics, at 1200 °C a two-layered oxide film is formed. The upper layer is comprised of amorphous Hf-doped oxide layer, and the lower layer is crystalline crystoballite α-SiO2. TEM investigation of samples oxidized at 1650 °C revealed the formation of HfSiO4 precipitates in the α-SiO2 matrix. In the case of the TLS ceramics, regardless of oxidation temperature two-layered oxide film consists of upper SiO2 and lower HfSiO4. TLS ceramics demonstrated the highest oxidation resistance under the plasma torch and kept its structural integrity during 180 s at 2000 °C.

New Permeable Structure for Controlling Debris Flows in the Wenjiagou Gully

Aimed to control debris flows more effectively, we attempt to develop a new structure called a “sloping roof grill dam”, which consists of guidance walls, separation grills and lateral deposit areas. A series of flume tests were performed to verify the function of this new structure, focusing on the trapping efficiency (β), deceleration efficiency (μ) and coarse-fine sediment separation extents (η) under different conditions of structure and debris flows. Experimental results show that both the trapping and deceleration efficiency increase with a decrease in permeability, whereas the separation extent shown the opposite effect. Multiple regression analysis indicates that β and η are linearly dependent on the permeability, deceleration efficiency and magnitudes of debris flow. In addition, an increase in beam gap can lead to an increase in the impact area downstream for the debris flow after the control of the structure. Finally, a two-level installation of the structure is tested to determine whether there is an enhancement of the debris control, which indicated that although the trapping effect of the two-level structure is considerably greater than that of a single-level structure (2.11 times at maximum), the separation extent of the two-level structure is not distinct and may be even lower.

Simulation modelling and analysis of dispatching rules in an assembly job shop production system with machine breakdowns

This paper addresses the scheduling problem in assembly job shop systems with machine breakdowns. The present study intends to rank the dispatching rules that are applied in an assembly job shop production system. A simulation model of an assembly job shop is developed for the purpose of this ranking and eight dispatching rules from the literature are incorporated in the simulation model. The product structures considered in this study are single-level assembly structure, two-level assembly structures and three level assembly structures. The machines are subjected to non-availability due to of breakdowns. Five performance measures are considered for analysis. The performance of each dispatching rule for each performance measure is calculated individually. To identify which of the dispatching rule provides the optimum result when all the performance measures are equally important, grey relational analysis is adopted to rank the dispatching rules. The results indicate that earliest completion time rule performs better in comparison with the other rules investigated in this study for single-level structure with multiple objectives.

A deep structure for human pose estimation

Articulated human pose estimation in unconstrained conditions is a great challenge. We propose a deep structure that represents a human body in different granularity from coarse-to-fine for better detecting parts and describing spatial constrains between different parts. Typical approaches for this problem just utilize a single level structure, which is difficult to capture various body appearances and hard to model high-order part dependencies. In this paper, we build a three layer Markov network to model the body structure that separates the whole body to poselets (combined parts) then to parts representing joints. Parts at different levels are connected through a parent–child relationship to represent high-order spatial relationships. Unlike other multi-layer models, our approach explores more reasonable granularity for part detection and sophisticatedly designs part connections to model body configurations more effectively. Moreover, each part in our model contains different types so as to capture a wide range of pose modes. And our model is a tree structure, which can be trained jointly and favors exact inference. Extensive experimental results on two challenging datasets show the performance of our model improving or being on-par with state-of-the-art approaches.

A hierarchical intrusion detection model based on the PCA neural networks

Most of existing intrusion detection (ID) models with a single-level structure can only detect either misuse or anomaly attacks. A hierarchical ID model using principal component analysis (PCA) neural networks is proposed to overcome such shortages. In the proposed model, PCA is applied for classification and neural networks are used for online computing. Experimental results and comparative studies based on the 1998 DARPA evaluation data sets are given, which show the proposed model can classify the network connections with satisfying performance.

RANKING STOCKS USING THE FUZZY MULTIPLE CRITERIA DECISION MAKING APPROACH

This work applies the Fuzzy Multiple Criteria Decision Making (FMCDM) approach to assist in making stock investment decisions. The proposed approach applies to quantitative data and can accommodate qualitative information which is normally difficult to be integrated by traditional finance and accounting methods. A major-sub-criteria hierarchy is established to reduce the possibility of over-weighing some dependent criteria existing in a single-level structure. The ratings of stocks versus qualitative sub-criteria and the weights of major and sub-criteria are assessed in linguistic terms represented by fuzzy numbers. Each sub-criterion is in a benefit, cost, or balanced nature. New standardization methods for cost-nature and balanced-nature criteria are presented. The algorithms of membership functions of the final aggregation are derived from the roots of cubic equations of multiplications of triple fuzzy numbers. Since these algorithms are clearly developed, the investor can easily calculate the fuzzy aggregation values. The defuzzified final aggregation judges the performance of alternative stocks. Moreover, the ratio of the market price to performance (PP) is suggested to filter the over and under-pricing of alternative stocks. A set of buying and selling rules are recommended based on the performance and PP ratio. Finally, an empirical example of evaluating a set of TSE-listed stocks tests the proposed approach.

Intrusion detection using hierarchical neural networks

Most intrusion detection system (IDS) with a single-level structure can only detect either misuse or anomaly attacks. Some IDSs with multi-level structure or multi-classifier are proposed to detect both attacks, but they are limited in adaptively learning. In this paper, two hierarchical IDS frameworks using Radial Basis Functions (RBF) are proposed. A serial hierarchical IDS (SHIDS) is proposed to identify misuse attack accurately and anomaly attacks adaptively. A parallel hierarchical IDS (PHIDS) is proposed to enhance the SHIDS’s functionalities and performance. The experiments show that the two proposed IDSs can detect network intrusions in real-time, train new classifiers for novel intrusions automatically, and modify their structures adaptively after new classifiers are trained.

Thermal modeling and management in ultrathin chip stack technology

This paper presents a thermal modeling for power management of a new three-dimensional (3-D) thinned dies stacking process. Besides the high concentration of power dissipating sources, which is the direct consequence of the very interesting integration efficiency increase, this new ultra-compact packaging technology can suffer of the poor thermal conductivity (about 700 times smaller than silicon one) of the benzocyclobutene (BCB) used as both adhesive and planarization layers in each level of the stack. Thermal simulation was conducted using three-dimensional (3-D) FEM tool to analyze the specific behaviors in such stacked structure and to optimize the design rules. This study first describes the heat transfer limitation through the vertical path by examining particularly the case of the high dissipating sources under small area. First results of characterization in transient regime by means of dedicated test device mounted in single level structure are presented. For the design optimization, the thermal draining capabilities of a copper grid or full copper plate embedded in the intermediate layer of stacked structure are evaluated as a function of the technological parameters and the physical properties. It is shown an interest for the transverse heat extraction under the buffer devices dissipating most the power and generally localized in the peripheral zone, and for the temperature uniformization, by heat spreading mechanism, in the localized regions where the attachment of the thin die is altered. Finally, all conclusions of this analysis are used for the quantitative projections of the thermal performance of a first demonstrator based on a three-levels stacking structure for space application.

Multilevel fuzzy relational systems: structure and identification

Existing fuzzy relational equations (FRE) typically possess an evident single-level structure, where no consequence part of the rule being modeled, is used as a fact to another rule. Corresponding to multistage fuzzy reasoning, a natural extension of traditional fuzzy relational systems (FRS) is to introduce some intermediate levels of processing governed by enhanced FRE's so that the structure resulted becomes multilevel or multistage. Three basic multilevel FRS structures, namely, incremental, aggregated, and cascaded, are considered in this paper and they correspond to different reasoning mechanisms being frequently used by human beings in daily life. While the research works on multilevel FRS are sparse and our ability to solve a system of multilevel FRE's in a purely analytical manner is very limited, we address the identification problem from an optimization approach and introduce three fuzzy neural models. The proposed models consist of single-level FRS modules that are arranged in different hierarchical manners. Each module can be realized by Lin and Lee's fuzzy neural model for implementing the Mamdani fuzzy inference. We have particularly addressed the problem of how to distribute the input variables to different (levels of) relational modules for the incremental and aggregated models. In addition, the new models can learn a complete multistage fuzzy rule set from stipulated data pairs using structural and parameter learning. The effectiveness of the multilevel models has been demonstrated through various benchmarking problems. It can be generally concluded that the new models are distinctive in learning, generalization, and robustness.

High resolution Gamow-teller strength measurement via β-decay of 65Ge

The β + decay of 65Ge was studied by β-delayed proton and γ-ray emission. The He-jet technique was used. Altogether 28 proton groups (1.0 MeV< E p<2.1 MeV) were observed with proton energy resolution of about 10 keV. The proton spectrum apparently consists of discrete proton groups due to a selective population of excited states of 65Ga by Gamow-Teller β + decay. The total proton branching ratio was measured to be (1.1±0.3)×10 −4. Upper limit of log ft values for beta transitions associated with the observed delayed proton groups and γ-rays were calculated and an experimental beta transition strength function was obtained. The single level structure of 65Ga was studied by the 64Zn(p, γ) 65Ga resonance reaction in the proton energy range 1.1 MeV< E p<2.1 MeV with 1.0 keV overall energy resolution and about 170 resonance levels were observed. Angular distribution and resonance decay measurements were done to study the nature of selected single levels and decay properties of low energy levels. The proton binding energy of 65Ga was measured to be B p = 3943.0±1.0 keV.