Two-step Optimization Strategy Research Articles

Late Fusion Multiview Clustering via Min-Max Optimization.

Multiview clustering (MVC) sufficiently exploits the diverse and complementary information among different views to improve the clustering performance. As a representative algorithm of MVC, the newly proposed simple multiple kernel k -means (SimpleMKKM) algorithm takes a min-max formulation and applies a gradient descent algorithm to decrease the resultant objective function. It is empirically observed that its superiority is attributed to the novel min-max formulation and the new optimization. In this article, we propose to integrate the min-max learning paradigm adopted by SimpleMKKM into late fusion MVC (LF-MVC). This leads to a tri-level max-min-max optimization problem with respect to the perturbation matrices, weight coefficient, and clustering partition matrix. To solve this intractable max-min-max optimization problem, we design an efficient two-step alternative optimization strategy. Furthermore, we analyze the generalization clustering performance of the proposed algorithm from the theoretical perspective. Comprehensive experiments have been conducted to evaluate the proposed algorithm in terms of clustering accuracy (ACC), computation time, convergence, as well as the evolution of the learned consensus clustering matrix, clustering with different numbers of samples, and analysis of the learned kernel weight. The experimental results show that the proposed algorithm is able to significantly reduce the computation time and improve the clustering ACC when compared to several state-of-the-art LF-MVC algorithms. The code of this work is publicly released at: https://xinwangliu.github.io/Under-Review.

Task-Free Continual Generation and Representation Learning via Dynamic Expansionable Memory Cluster

Human brains can continually acquire and learn new skills and knowledge over time from a dynamically changing environment without forgetting previously learnt information. Such a capacity can selectively transfer some important and recently seen information to the persistent knowledge regions of the brain. Inspired by this intuition, we propose a new memory-based approach for image reconstruction and generation in continual learning, consisting of a temporary and evolving memory, with two different storage strategies, corresponding to the temporary and permanent memorisation. The temporary memory aims to preserve up-to-date information while the evolving memory can dynamically increase its capacity in order to preserve permanent knowledge information. This is achieved by the proposed memory expansion mechanism that selectively transfers those data samples deemed as important from the temporary memory to new clusters defined within the evolved memory according to an information novelty criterion. Such a mechanism promotes the knowledge diversity among clusters in the evolved memory, resulting in capturing more diverse information by using a compact memory capacity. Furthermore, we propose a two-step optimization strategy for training a Variational Autoencoder (VAE) to implement generation and representation learning tasks, which updates the generator and inference models separately using two optimisation paths. This approach leads to a better trade-off between generation and reconstruction performance. We show empirically and theoretically that the proposed approach can learn meaningful latent representations while generating diverse images from different domains. The source code and supplementary material (SM) are available at https://github.com/dtuzi123/DEMC.

A multifidelity Bayesian optimization method for inertial confinement fusion design

Due to their cost, experiments for inertial confinement fusion (ICF) heavily rely on numerical simulations to guide design. As simulation technology progresses, so too can the fidelity of models used to plan for new experiments. However, these high-fidelity models are by themselves insufficient for optimal experimental design, because their computational cost remains too high to efficiently and effectively explore the numerous parameters required to describe a typical experiment. Therefore, traditionally, ICF design has relied on low-fidelity modeling to initially identify potentially interesting design regions, which are then subsequently explored via selected high-fidelity modeling. In this paper, we demonstrate that this two-step approach can be insufficient: even for simple design problems, a two-step optimization strategy can lead high-fidelity searching toward incorrect regions and consequently waste computational resources on parameter regimes far away from the true optimal solution. We reveal that a primary cause of this behavior in ICF design problems is the presence of low-fidelity optima in different regions of the parameter space far away from high-fidelity optima. To address this issue, we propose an iterative multifidelity Bayesian optimization method based on Gaussian Process Regression that leverages both low- and high-fidelity models simultaneously. We demonstrate, using both two- and eight-dimensional ICF test problems, that our algorithm can effectively utilize both low-fidelity and high-fidelity models to refine the designs. This approach proves to be more efficient than relying solely on high-fidelity modeling for optimization.

Design and optimization of 3D-bioprinted cell-laden scaffolds in dynamic culture

Light-based 3D printing enables the fabrication of biological scaffolds with high precision, versatility and biocompatibility, particularly the cell-laden scaffolds with architecturally complex geometric features. However, many bioprinted tissue scaffolds suffer from low cell viability due to insufficient oxygen and nutrient supply, which is heavily influenced by scaffold structure and cultivation conditions. Current practice relies mainly on resource-intensive trial-and-error methods to optimize scaffolds’ structures and cultivation parameters. In this study, we developed a comprehensive multi-physics model integrating fluid dynamics, oxygen mass transfer, cell oxygen consumption, and cell growth processes to capture cell growth behaviors in scaffolds, establishing a robust theoretical foundation for scaffold structure optimization. The modeling results showed that a large number of parameters, such as system inlet flow rate, geometric feature size, cell parameters, and material properties, significantly impact oxygen concentration and cell growth within the scaffold. A two-step optimization strategy is proposed in this paper and was applied to obtain optimal geometric parameters of channeled scaffolds to demonstrate the model’s effectiveness for scaffold optimization. The model can be employed for scaffolds with arbitrary shapes and various materials, facilitating the optimal design of sophisticated scaffolds for more advanced tissue engineering.

An analytical target cascading method-based two-step distributed optimization strategy for energy sharing in a virtual power plant

Large-scale distributed renewable energy sources as well as emerging controllable loads such as electric vehicles connected to the distribution grids have posed challenges to the efficient management of these distributed energy resources. Promoting energy-sharing of distributed energy resources through economic incentives is a cost-effective and equitable solution to the challenge. This paper proposes a two-stage transactive energy control mechanism by the virtual power plant (VPP) for energy-sharing of multiple prosumers. Firstly, a VPP internal price mechanism based on the supply-demand ratio of prosumers in the energy-sharing alliance is proposed. Then, a two-stage economic optimization model is developed. Trading prices between VPP and prosumers are established in the first stage, and prosumers make operating plans based on those prices in the second stage. Then, considering that prosumers and VPP are independent stakeholders, a decentralized optimization algorithm based on the analytical target cascading method is developed. The upper-level VPP coordinates the energy-sharing alliance and sets the transaction price following the overall supply and demand ratio. Lower-level prosumers independently make decisions and feedback on the trading power of the VPP. Then, iterative calculations provide a distributed and independent solution. Finally, the case study verifies that the proposed energy-sharing model can reduce the operating cost of prosumers by 14.29 %. Moreover, prosumer privacy protection is achieved by the analytical target cascading-based distributed model at a cost of less than 1.5 % accuracy.

Feature-constrained real-time simultaneous monitoring of monocular vision odometry for bridge bearing displacement and rotation

The displacement and rotation are critical indicators of bridge's health, especially for the key components of the bridge such as bridge bearings. This study presents a real-time simultaneous measurement method of bridge bearing displacement and rotation to realize round-the-clock structure monitoring. The main contributions include: (1) A displacement and rotation simultaneous measurement algorithm and system is developed based on feature-constrained monocular visual odometry (VO), which uses image sequences of infrared array lamp targets with multiple feature constraints as visual input to construct the mathematical relationship between displacement and rotation and coordinate system transformation matrix. (2) A two-step parameter optimization strategy is proposed based on adaptive multi-window centroid array tracking and transformation matrix purification and refinement, effectively improving displacement and rotation measurement accuracy and robustness. The effectiveness of the proposed method is verified in the laboratory experiment and the actual bridge bearing monitoring of the Huangpu Pearl River Bridge.

Synergistic band modulation and precipitates: Achieving high quality factor in SnTe

Breaking the thermoelectric figure of merit zT barrier of SnTe enables it to become a promising alternative to PbTe; however, the inferior and strongly coupled physicochemical properties of pristine SnTe severely restrict the efficient optimization. Herein, we doped trivalent Sb in SnTe and incorporated SnS particles to achieve high quality factor B through a two-step optimization strategy of tuning the valence band structure and intercalating heterostructural precipitates, and well predicted the potential prospects. The high solubility limit of Sb not only reduced the carrier concentration nH but also significantly optimized the valence band structure and improved the Seebeck coefficient, thereby enhancing the weight mobility μw in the all-temperature region. Furthermore, the additional SnS, which tends to exist as precipitates with different micrometer-scale sizes, enhanced low-medium-frequency phonon scattering in a wider frequency range except for point defects scattering, suppressing the lattice thermal conductivity to 0.55 W m−1 K−1. As a result of this synergistic effect, a high B-factor of ∼0.82 greater than triple pure SnTe was obtained in Sn0.91Sb0.09Te-10%SnS, with an enhanced zT of ∼1.15 at 850 K. More importantly, the high B-factor accurately predicted an excellent zT value of ∼1.65 at the optimal Fermi level, which highlights the great potential of Sn1-xSbxTe-y%SnS-based materials. This work provides an effective route for stepwise optimization of electrical and thermal performance from the B-factor perspective and has guiding significance for other thermoelectric materials.

Verification of improved YOLOX model in detection of greenhouse crop organs: Considering tomato as example

Efficient and accurate monitoring crop growth is an important prerequisite for modern greenhouse crop growth control and management. However, it is still challenging to achieve efficient in coordinated detection of tomato organs. Therefore, in this study, a multiobjective detection model was established to detect the blossom, fruit, and fruit color of crop population under dense planting, leaf shading, fruit overlap, and interference from light change in greenhouses. Specifically, sample equalization, random transformation, and mosaic enhancement were applied to identify multiple targets and small-sample images. A new detection network, YOLOX, was used, which introduces a decoupled head structure and redesigns a detection head according to the anchor-free concept to effectively improve the recognition of small-target occlusion. An efficient tomato flower and fruit recognition model was constructed based on a two-step fine-tuned multi-round optimization strategy training. The results showed that the YOLOX target detection model improved the average recognition accuracy (mAP) by 52.49–10.75 %, was lighter, was only 14 % original YOLOv4, and improved the detection speed by 25.53 FPS compared to six main target detection algorithms. It can be used as a method for the collaborative recognition of tomato flowers and fruits. Accordingly, two sets of lightweight detection networks with different performances were designed according to different agricultural scenarios. YOLOXMOB modeled with a small size (only 34.2 MB), high mAP of 81.39 %, and detection speed of 61.85 FPS. It is recommended for use as a deployment model for mobile devices. YOLOXPC, with a deep network layer and strong performance, achieved 86.16 % mAP, and is recommended for monitoring and dispatching computers. This study puts forward a set of multiobjective detection model scheme adapted to the real environment of greenhouse, which provides ideas for the construction of a highly intelligent monitoring system platform for the agricultural Internet of Things.

Multivariable optimization process of heterotrophic growth of Chlorella vulgaris

Microalgae have received increasing attention as one of the most promising feedstocks in the development of new healthier food products and different strategies have been attempted to improve their growth. However, the high production costs and low productivities, commonly associated with photoautotrophic growths, are still a big challenge. In this study, a two-step optimization strategy was carried out in order to maximize the biomass production of a Chlorella vulgaris strain used at industrial scale under heterotrophic conditions. From a total of 24 independent variables, which were studied simultaneously, 10 have presented a positive effect over Xmax, while the remaining have shown to be negative. The amount of (NH4)2SO4 (6.3 g L−1), MgSO4·7H2O (0.7 g L−1), and C6H12O6 (50% w/v) in the culture medium has revealed to be the only factors with a significant impact on biomass concentration, with optimum values of 25.5, 64.6, and 75 ml.L−1, respectively. The optimized medium resulted in an improvement of the Xmax by 99.6% when compared to the growth medium applied at industrial scale, proving the success of this strategy. Additionally, the carbohydrates production was enhanced by 48.0%.

Rethinking adversarial domain adaptation: Orthogonal decomposition for unsupervised domain adaptation in medical image segmentation.

Medical image segmentation methods based on deep learning have made remarkable progress. However, such existing methods are sensitive to data distribution. Therefore, slight domain shifts will cause a decline of performance in practical applications. To relieve this problem, many domain adaptation methods learn domain-invariant representations by alignment or adversarial training whereas ignoring domain-specific representations. In response to this issue, this paper rethinks the traditional domain adaptation framework and proposes a novel orthogonal decomposition adversarial domain adaptation (ODADA) architecture for medical image segmentation. The main idea behind our proposed ODADA model is to decompose the input features into domain-invariant and domain-specific representations and then use the newly designed orthogonal loss function to encourage their independence. Furthermore, we propose a two-step optimization strategy to extract domain-invariant representations by separating domain-specific representations, fighting the performance degradation caused by domain shifts. Encouragingly, the proposed ODADA framework is plug-and-play and can replace the traditional adversarial domain adaptation module. The proposed method has consistently demonstrated effectiveness through comprehensive experiments on three publicly available datasets, including cross-site prostate segmentation dataset, cross-site COVID-19 lesion segmentation dataset, and cross-modality cardiac segmentation dataset. The source code is available at https://github.com/YonghengSun1997/ODADA.

Synthesizing Circularly Polarized Multi-Beam Planar Dipole Arrays With Sidelobe and Cross-Polarization Control by Two-Step Element Rotation and Phase Optimization

In this article, a novel method is presented to synthesize multi-beam circularly polarized (CP) patterns with sidelobe level (SLL) and cross-polarization level (XPL) suppression for planar dipole arrays by optimizing both element rotations and excitation phases. Multiple sets of excitation phases are adopted as optimization variables other than the determined values according to the rotation angles in the conventional sequentially rotated technique (SRT) and the random sequentially rotated technique (RSRT), so as to improve the performance of the obtained CP multi-beam patterns. This releases many more degrees of synthesis freedom but leads to a new problem of demanding excessive computation resources for a planar rotated dipole array with multiple CP patterns. Then, a two-step rotation and phase optimization strategy is proposed to effectively solve this optimization problem. In the initial step, the common element rotations and approximated excitation phases composed of the common phase parts plus the beam focusing phase parts are found for multiple desired CP patterns with different beam directions. And then, a refining step is performed to refine the excitation phases individually for each CP beam for improved pattern performance. Four representative examples of synthesizing multi-beam patterns with both SLL and XPL suppression are provided to verify the effectiveness of the proposed method. Comparisons with the conventional SRT and RSRT are also provided to show the superiority of the proposed method.

Design of a Hybrid Metal-Plasma Transmit-Array With Beam-Scanning Capabilities

This article presents the complete design of a hybrid metal-plasma transmit-array antenna with beam-scanning capabilities. The antenna operates at 1.07 GHz and consists of an active metallic dipole, a metallic ground plane (GP), and a set of cylindrical plasma discharges arranged in a 2-D lattice. The main lobe of the antenna can be tilted toward different directions simply by turning on/off specific subsets of plasma discharges. Toward this aim, a two-step optimization strategy based on the combination of a particle swarm optimization (PSO) and of a genetic algorithm (GA) is adopted. A realistic model of the antenna has been simulated in CST Microwave Studio. To derive plasma parameters, a prototype plasma discharge has been realized and tested. Each plasma bar included in the numerical setup comprises not only the plasma material, but also the glass envelope and the metallic electrodes used to confine and generate the plasma, respectively. An analysis has been carried out to evaluate the effects of adding the electrically long metallic cables to each discharge needed to ignite and sustain the plasma. The radiation properties of the array can be maintained thanks to the insertion of inductances between the electrodes and the cables. Finally, a sensitivity analysis has been conducted to identify intervals of plasma density and neutral gas pressure within which the beam-scanning capabilities of the antenna remain unaffected.

Robustness Improvement of Using Pre-Trained Network in Visual Odometry for On-Road Driving

Robustness in on-road driving Visual Odometry (VO) systems is critical, as it determines the reliable performance in various scenarios and environments. Especially with the development of data-driven technology, the combination of data-driven VO and model-based VO has achieved accurate tracking performance. However, the lack of generalization of pre-trained deep neural networks (DNN) limits the robustness of such a combination in unseen environments. In this study, we introduce a novel framework with appropriate usage of DNN prediction and improve the robustness in the self-driving application. Based on the characteristic of on-road self-driving motion and the DNN output, we propose a two-step optimization strategy with a variable degree of freedom (DoF), i.e., the use of two types of DoF representations during pose estimation. Specifically, our two-step optimization operates according to the residual of the optimization with the motion label classification from the pre-trained DNN, as well as our proposed Motion Evaluation by essential matrix construction. Experimental results show that our framework obtains better tracking accuracy than the existing methods.

A Two-Step Optimization for Extrinsic Calibration of Multiple Camera System (MCS) Using Depth-Weighted Normalized Points

A multiple camera system (MCS), which allows only a limited common field-of-view between adjacent cameras, has been chiefly calibrated using a typical 2D-3D point correspondence. However, the correspondence contains a potential instability by which the calibration result can diverge, and the instability has not been studied nor overcome before. We propose a MCS extrinsic calibration method with high robustness and accuracy based on a two-step optimization strategy. Using depth-weighted normalized points, we develop two novel types of point correspondence as follows. The 1st correspondence, F <sub xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">cb</sub> , aims to robustly estimate the MCS extrinsic parameters by overcoming the potential instability that exists in the typical 2D-3D point correspondence. The 2nd correspondence, F <sub xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">cc</sub> , aims to refine the MCS extrinsic parameters by using the direct relation between adjacent cameras. In the simulation, we validated the robustness and high accuracy of the proposed method. We validated its high precision in the field test. In both the simulation and the field test, our method was compared with the state-of-the-art method.

Parametric automatic optimal design of USV hull form with respect to wave resistance and seakeeping

As the Unmanned Surface Vehicle (USV) plays an increasingly important role in ocean exploration, the performance requirements of the USV are becoming higher and higher. This paper presents a parametric automatic multi-objective optimization design of USV based on two-step optimization strategy. The focus of our work is on the parametric automatic optimization design of USV hull form with respect to resistance and seakeeping in waves. The wave resistance, the vertical acceleration of the center of gravity, and the pitch amplitude of the USV are selected as the optimization objectives. The parametric modeling and design variables are selected for the hull. Second order Response Surface Method (RSM) is used to build approximate model instead of numerical simulation in the optimization process to reduce the optimization time. The Particle Swarm Optimization (PSO) algorithm and Sequential Quadratic Programming (SQP) algorithm are used to solve the optimization model. Through the integration of NX, STAR-CCM+, and Excel in the intelligent platform, automatic optimization is realized. The effectiveness of the automatic optimization design method is verified by taking a 7 m USV as an example. This method provides a solution for the automatic optimization design of the USV hull form.

Content-Based Visual Summarization for Image Collections.

With the surge of images in the information era, people demand an effective and accurate way to access meaningful visual information. Accordingly, effective and accurate communication of information has become indispensable. In this article, we propose a content-based approach that automatically generates a clear and informative visual summarization based on design principles and cognitive psychology to represent image collections. We first introduce a novel method to make representative and nonredundant summarizations of image collections, thereby ensuring data cleanliness and emphasizing important information. Then, we propose a tree-based algorithm with a two-step optimization strategy to generate the final layout that operates as follows: (1) an initial layout is created by constructing a tree randomly based on the grouping results of the input image set; (2) the layout is refined through a coarse adjustment in a greedy manner, followed by gradient back propagation drawing on the training procedure of neural networks. We demonstrate the usefulness and effectiveness of our method via extensive experimental results and user studies. Our visual summarization algorithm can precisely and efficiently capture the main content of image collections better than alternative methods or commercial tools.

Integrated design of topology and material for composite morphing trailing edge based compliant mechanism

The morphing trailing edge based compliant mechanism is a developing technology which can increase lift-drag ratio for variable flight modes by bending down the trailing edge. Composite material design is integrated into topology optimization for the morphing trailing edge based compliant mechanism in the paper. A two-step optimization strategy is established to solve the integrated design problem. Initially, lamination parameters are introduced and viewed as a bridge between structure stiffness and fiber angles for composite material. Design variables include the lamination parameters and element density. The least-squares between actual and desired displacements at output points along trailing edge is adopted to evaluate the deformed capability of the trailing edge. An integrated optimization model for the composite morphing trailing edge is established with the volume constraints. The optimal topologic shape and lamination parameters are initially obtained. Subsequently, a least-squares optimization between fiber angles and the optimal lamination parameters is implemented to obtain optimal fiber angles. Finally, morphing capability of composites trailing edge based compliant mechanism is investigated by simulation and experiments. The results indicate the composites trailing edge based compliant mechanism can approximately bend down 8 degrees and satisfies the design requirement.

Highly Converged Valence Bands and Ultralow Lattice Thermal Conductivity for High-Performance SnTe Thermoelectrics.

A two-step optimization strategy is used to improve the thermoelectric performance of SnTe via modulating the electronic structure and phonon transport. The electrical transport of self-compensated SnTe (that is, Sn1.03 Te) was first optimized by Ag doping, which resulted in an optimized carrier concentration. Subsequently, Mn doping in Sn1.03-x Agx Te resulted in highly converged valence bands, which improved the Seebeck coefficient. The energy gap between the light and heavy hole bands, i.e. ΔEv decreases to 0.10 eV in Sn0.83 Ag0.03 Mn0.17 Te compared to the value of 0.35 eV in pristine SnTe. As a result, a high power factor of ca. 24.8 μW cm-1 K-2 at 816 K in Sn0.83 Ag0.03 Mn0.17 Te was attained. The lattice thermal conductivity of Sn0.83 Ag0.03 Mn0.17 Te reached to an ultralow value (ca. 0.3 W m-1 K-1 ) at 865 K, owing to the formation of Ag7 Te4 nanoprecipitates in SnTe matrix. A high thermoelectric figure of merit (z T≈1.45 at 865 K) was obtained in Sn0.83 Ag0.03 Mn0.17 Te.

Automatic Strain Gauge Balance Design Optimization Approach and Implementation Based on Integration of Software

Abstract The traditional wind tunnel strain balance design cycle is a manual iterative process. With the experience and intuition of the designer, one solution that meets the design requirements can be selected among a small number of design solutions. This paper introduces a novel software integration-based automatic balance design optimization system (ABDOS) and its implementation by integrating professional design knowledge and experience, stepwise optimization strategy, CAD-CAE software, self-developed scripts and tools. The proposed two-step optimization strategy includes the analytical design process (ADP) and the finite element method design process (FEDP). The built-in optimization algorithm drives the design variables change and searches for the optimal structure combination meeting the design objectives. The client-server based network architecture enables local lightweight design input, task management, and result output. The high-performance server combines all design resources to perform all the solution calculations. The development of more than 10 balances that have been completed and a case study show that this method and platform significantly reduce the time for design evaluation and design-analysis-redesign cycles, assisting designers to comprehensively evaluate and improve the performance of the balance.

Optimal orbit design for repeat-pass multi-baseline observation of tomographic SAR satellites

Tomographic synthetic aperture radar (TomoSAR) is a modern technique that allows retrieving the structural information of detected objects. This paper investigates the problem of optimal orbit design in TomoSAR space missions. Considering the mission requirement of multi-baseline repeated observation, the orbital characteristics of TomoSAR satellites are analysed under the J2 perturbation effect. By introducing the performance indicators of sampling period and orbital transfer fuel consumption, the orbit design is transformed into a multi-objective optimization problem. A two-step optimization strategy is developed to find the optimal orbits of TomoSAR satellites. In the first step, where some approximations are made, the idea of Pareto optimal solutions is applied under the objectives of time and fuel. A more accurate model is utilized in the second refinement step, and biogeography-based optimization (BBO) is employed to obtain satisfactory revisit accuracy. Numerical simulations are conducted to testify the effectiveness of the suggested design process, providing references for TomoSAR missions in terms of satellite orbit.