Iterative Optimization Research Articles

RNA sample optimization for cryo-EM analysis.

RNAs play critical roles in most biological processes. Although the three-dimensional (3D) structures of RNAs primarily determine their functions, it remains challenging to experimentally determine these 3D structures due to their conformational heterogeneity and intrinsic dynamics. Cryogenic electron microscopy (cryo-EM) has recently played an emerging role in resolving dynamic conformational changes and understanding structure-function relationships of RNAs including ribozymes, riboswitches and bacterial and viral noncoding RNAs. A variety of methods and pipelines have been developed to facilitate cryo-EM structure determination of challenging RNA targets with small molecular weights at subnanometer to near-atomic resolutions. While a wide range of conditions have been used to prepare RNAs for cryo-EM analysis, correlations between the variables in these conditions and cryo-EM visualizations and reconstructions remain underexplored, which continue to hinder optimizations of RNA samples for high-resolution cryo-EM structure determination. Here we present a protocol that describes rigorous screenings and iterative optimizations of RNA preparation conditions that facilitate cryo-EM structure determination, supplemented by cryo-EM data processing pipelines that resolve RNA dynamics and conformational changes and RNA modeling algorithms that generate atomic coordinates based on moderate- to high-resolution cryo-EM density maps. The current protocol is designed for users with basic skills and experience in RNA biochemistry, cryo-EM and RNA modeling. The expected time to carry out this protocol may range from 3 days to more than 3 weeks, depending on the many variables described in the protocol. For particularly challenging RNA targets, this protocol could also serve as a starting point for further optimizations.

RFL-LSU: A Robust Federated Learning Approach with Localized Stepwise Updates

Distributed intelligence enables the widespread deployment of AI technology, greatly promoting the development of AI. Federated learning is a widely used distributed intelligence technology that allows iterative optimization of global model while protecting user data privacy. Currently, federated learning faces some security threats, as its open architecture provides attackers with opportunities to disrupt the learning process by submitting malicious updates or inserting backdoors. In this article, we propose a robust federated learning method to defend against potential malicious attacks. Specifically, we enhance the algorithm’s performance and stability by implementing localized stepwise updates on the client side and element-wise anomaly detection on the server side. We conducted experiments in a more realistic non-i.i.d. scenario and compared the results with other typical federated learning methods. Our results demonstrate that our approach exhibits strong robustness under non-i.i.d. data distributions, outperforming other methods in terms of test accuracy and resilience to attacks.

Stochastic Gradient Descent with Pre-Conditioned Polyak Step-Size

Stochastic Gradient Descent (SGD) is one of the many iterative optimization methods that are widely used in solving machine learning problems. These methods display valuable properties and attract researchers and industrial machine learning engineers with their simplicity. However, one of the weaknesses of this type of methods is the necessity to tune learning rate (step-size) for every loss function and dataset combination to solve an optimization problem and get an efficient performance in a given time budget. Stochastic Gradient Descent with Polyak Step-size (SPS) is a method that offers an update rule that alleviates the need of fine-tuning the learning rate of an optimizer. In this paper, we propose an extension of SPS that employs preconditioning techniques, such as Hutchinson’s method, Adam, and AdaGrad, to improve its performance on badly scaled and/or ill-conditioned datasets.

An integrated optimization strategy for mission planning and control resource allocation for long-endurance unmanned aerial vehicle

PurposeWith burgeoning interest in the low-altitude economy, applications of long-endurance unmanned aerial vehicles (LE-UAVs) have increased in remote logistics distribution. Given LE-UAVs’ advantages of wide coverage, strong versatility and low cost, in addition to logistics distribution, they are widely used in military reconnaissance, communication relay, disaster monitoring and other activities. With limited autonomous intelligence, LE-UAVs require regular periodic and non-periodic control from ground control resources (GCRs) during flights and mission execution. However, the lack of GCRs significantly restricts the applications of LE-UAVs in parallel.Design/methodology/approachWe consider the constraints of GCRs, investigating an integrated optimization problem of multi-LE-UAV mission planning and GCR allocation (Multi-U&G IOP). The problem integrates GCR allocation into traditional multi-UAV cooperative mission planning. The coupling decision of mission planning and GCR allocation enlarges the decision space and adds complexities to the problem’s structure. Through characterizing the problem, this study establishes a mixed integer linear programming (MILP) model for the integrated optimization problem. To solve the problem, we develop a three-stage iterative optimization algorithm combining a hybrid genetic algorithm with local search-variable neighborhood decent, heuristic conflict elimination and post-optimization of GCR allocation.FindingsNumerical experimental results show that our developed algorithm can solve the problem efficiently and exceeds the solution performance of the solver CPLEX. For small-scale instances, our algorithm can obtain optimal solutions in less time than CPLEX. For large-scale instances, our algorithm produces better results in one hour than CPLEX does. Implementing our approach allows efficient coordination of multiple UAVs, enabling faster mission completion with a minimal number of GCRs.Originality/valueDrawing on the interplay between LE-UAVs and GCRs and considering the practical applications of LE-UAVs, we propose the Multi-U&G IOP problem. We formulate this problem as a MILP model aiming to minimize the maximum task completion time (makespan). Furthermore, we present a relaxation model for this problem. To efficiently address the MILP model, we develop a three-stage iterative optimization algorithm. Subsequently, we verify the efficacy of our algorithm through extensive experimentation across various scenarios.

Unfolding Electrolyzer Characteristics to Reveal Solar-to-Chemical Efficiency Potential: Rapid Analysis Method Bridging Electrochemistry and Photovoltaics.

Development of photovoltaic-electrochemical (PV-EC) systems for energy storage and industry decarbonization requires multidisciplinary collaborative efforts of different research groups from both photovoltaic and electrochemical research communities. Consequently, the evaluation of the solar-to-chemical or solar-to-fuel efficiency of a new electrolyzer (EC) as a part of a PV-EC system is a time-consuming task that is challenging in a routine optimization loop. To address this issue, a new rapid assessment method is proposed. This method employs power balance requirements to unfold the input EC characteristics into the parameter space of PV-EC systems. The system parameters, composed with the EC output characteristics, yield the solar-to-chemical efficiency attainable by the electrolyzer in combination with any PV device under any irradiance at any relative PV-to-EC scaling and any mode of power coupling. This comprehensive overview is achieved via a mathematically simple conversion of the EC characteristics in any spreadsheet software. The method, designed to streamline the development and minimize the efforts of both the photovoltaic and electrochemical communities, is demonstrated via the analysis of CO2-reduction electrolyzer characteristics and verified with dedicated PV-EC experiments.

Multi-Objective Optimization of Three-Stage Turbomachine Rotor Based on Complex Transfer Matrix Method

This study presents the complex transfer matrix method (CTMM) as an advanced mathematical model, providing significant advantages over the finite element method (FEM) by yielding rapid solutions for complex optimization problems. In order to design a more efficient structure of a three-stage turbomachine rotor, we integrated this method with various optimization algorithms, including genetic algorithm (GA), differential evolution (DE), simulated annealing (SA), gravitational search algorithm (GSA), black hole (BH), particle swarm optimization (PSO), Harris hawk optimization (HHO), artificial bee colony (ABC), and non-metaheuristic pattern search (PS). Thus, the best rotor geometry can be obtained fast with minimum bearing forces and disk deflections within design limits. In the results, the efficiency of the CTMM for achieving optimized designs is demonstrated. The CTMM outperformed the FEM in both speed and applicability for complex rotordynamic problems. The CTMM was found to deliver results of comparable quality much faster than the FEM, especially with higher element quality. The use of the CTMM in the iterative optimization process is shown to be highly advantageous. Furthermore, it is noted that among the different optimization algorithms, ABC provided the best results for this multi-objective optimization problem.

Frigatebird optimizer: a novel metaheuristic algorithm

Abstract This paper proposes a new swarm intelligence optimization algorithm - the Frigate Bird Optimizer (FBO). The algorithm is inspired by the unique flight and foraging behaviors of frigatebirds. Its optimization process is divided into two stages: The first stage simulates frigatebirds harassing other seabirds to snatch food, and the search direction and radius of individuals have randomness and uncertainty, which is conducive to global exploration; The second stage simulates frigatebirds observing large fish driving small fish to leap out of the water and preying on them, and individuals tend to gather in the optimal search direction. By simulating the behaviors of frigatebirds under different survival strategies, the algorithm achieves extensive global search in the first stage and fine-tuned local optimization by learning information in the second stage. To evaluate its performance, 46 functions in the CEC2014 and CEC2017 benchmark test sets are selected as objective functions and compared with 9 state-of-the-art meta-heuristic algorithms. The results show that the FBO algorithm has higher performance, excellent iterative optimization ability and strong robustness, and can be applied to different optimization domains.

Targeted degradation of Pin1 by protein-destabilizing compounds

The concept of targeted protein degradation is at the forefront of modern drug discovery, which aims to eliminate disease-causing proteins using specific molecules. In this paper, we explored the idea to design protein degraders based on the section of ligands that cause protein destabilization, hence that facilitate the cellular breakdown of the target. Our studies present covalent agents targeting Pin1, a cis-trans prolyl isomerase that plays a crucial role in tumorigenesis. Our design strategy entailed iterative optimizations of agents for potency and Pin1 destabilization in vitro. Biophysical and cellular studies suggest that the agents may act like molecular crowbars, displacing protein-stabilizing interactions that open the structure for recognition by the proteasome degradation machinery. This approach resulted in a series of potent and effective Pin1 degraders with potential applications in target validation and in therapeutic development. We propose that our design strategy can identify molecular degraders without engineering bifunctional agents that artificially create interactions between a disease-causing protein and a ubiquitin ligase.

Shape design optimization of a flat endmill tool

AbstractAmong the many possible metal‐cutting processes, milling is regarded as one of the most widely applied processes to machine different engineering materials productively. During the milling process, there is relative motion between the endmill tool and the workpiece. Energy is required to drive the cutting force used by the tool in separating the chips from the workpiece. Several studies have focused on the effects of milling process parameters on the cutting force, for instance by varying the cutting parameters, machining properties and the extent of cooling/lubrication. Other studies have explored the influence of tool geometry and material properties on the cutting force. Results from these studies have brought about improvements in the design of milling machinery and tools. The current study presents a novel automated approach for tool geometry optimization which is fully coded in Matlab. An automated and parametrized endmill design procedure was created featuring input parameters such as rake angle, relief angle, clearance angle and helix angle, all of which have been shown, through numerous studies, to influence cutting forces. The parameters were here used to generate the shape profile for one cutting edge of a flat endmill. This cutting edge profile is next copied and rotated, as per the design's pitch angle, to form the complete set of cutting flutes for the flat endmill. Next, the surface defined by the completed profile is meshed using quadrilateral elements. The 3D endmill design is then formed by extruding, and twisting as per the helix angle, this quadrilateral mesh to form a solid hexahedral mesh. Finally, an automated iterative optimization process was defined, featuring the above parametrized design process coupled with Abaqus‐based finite element simulation of the milling process. The expected result is that the proposed optimization procedure will be able to identify the endmill design parameters which minimize the cutting force. At the completion of the optimization, a minimized maximum resultant cutting force value is obtained from using the newly optimized endmill. The resultant cutting force was reduced by 17.6%. It is anticipated that the automated design, meshing and simulation‐driven optimization approach is generalizable to other tool design variations.

Automated Virtual Machine Migration with Load Shifting Technique Avoiding Latency in Cloud Environment

Efficient management of Virtual Machines (VMs) in cloud computing is one key aspect to achieve the best utilization of resources and performance. The study is about improving Automated Virtual Machine Migration with a Load Shifting Technique in order to reduce latency through the cloud domain. This methodology dynamically distribute the load across multiple VMs by migrating their workloads to other physical hosts determined according certain resource utilization metrics available in real-time. The dynamic and unpredictable nature of cloud workloads requires a solution that is more responsive in the way it manages VMs. Live migration and pre-copy migration offer suboptimal performance in terms of time to migrate, downtime due to the prolonged copying period as heaps swell/collapse during live migrations. Utilizing sophisticated algorithms and real-time monitoring to avoid these pitfalls, our methodology allows us for proper load balancing with minimal latency. The primary elements of the proposed framework are a detailed monitoring module, an intelligent load shifting algorithm, high-performance migration method and feedback loop for continuous optimization. The monitoring module is responsible for collecting on-the-fly information about CPU, memory, network traffic and disk I/O such that can be exploited by the load shifting algorithm. The algorithm processes this data to decide which VMs should be migrated with the purpose of ensuring load balancing and decreasing latency achieved using these migrations. In order to test the proposed approach, comprehensive simulations and comparatives analyses utilizing other models were performed. This results in better latency reduction, increased resource utilization and improved overall system performance. In fact, the proposed method can handle not only current challenges in cloud computing such as load balancing and latency issued but also a scalable and effective approach for general cloud infrastructure usage. Finally, the Automated Virtual Machine Migration with Load Shifting Technique proves to be a reliable and successful solution for assisting in managing dynamic cloud workloads. Through real-time monitoring, advanced algorithms and continuous optimization this method has now been able to improve cloud services performance and reliability thereby allowing it to better serve service level agreements as well meet up with user satisfaction. We hope to adapt the algorithm and generalize it for even more advanced predictive models in future iterations, as well as evaluation of practical implementation on cloud-based deployment environments.

Cost Effectiveness of Chip Seal and Hot Mix Asphalt Pavements

Chip seal pavements, consisting of one or more layers of asphalt binder and fine aggregate, can be mechanically characterized as a surface treatment that enhances evenness and trafficability. This paper examines the geotechnical aspects of chip seal applicability compared to traditional hot mix asphalt pavements. An analytical model was employed to design unpaved roads and determine the required thickness of unbound layers. Eight optimization models were developed for hot mix asphalt pavements and four for chip seal pavements, aimed at achieving optimal designs for various input parameters. These outcomes were used to conduct a multi-parametric analysis, incorporating an optimization loop for each combination of design variables. The results indicate that, under low traffic conditions, a chip seal pavement structure can be up to 40% less expensive than an optimal hot mix asphalt pavement structure, particularly when the subgrade has low bearing capacity and is exposed to unfavorable climatic conditions. However, at medium traffic loads, with good subgrade bearing capacity and favorable climate, the chip seal pavement structure incurs costs that are 25% higher than those of the hot asphalt pavement structure. In addition, chip seal pavements should always be designed with integrated geosynthetic reinforcement to minimize construction costs, and chip seal is not as sensitive to frost as hot mix asphalt.

Planning and Evaluation of Water-Dropping Strategy for Fixed-Wing Fire Extinguisher Based on Multi-Resolution Modeling

The deployment of fixed-wing aircraft in fire-extinguishing operations represents a significant advancement in the domain of aviation emergency rescue. Addressing the challenge of enhancing firefighting efficacy, this study delves into the water-dropping strategies of fixed-wing extinguishers and provides a methodological framework for the strategic planning and assessment of water-dropping tactics, employing multi-resolution modeling. The formulation of the planning algorithm and the structure of the effectiveness evaluation index system are explained accordingly. The corresponding prototype system was designed, comprising four subsystems that utilized distinct resolution models: fire environment simulation, water-dropping point scheme planning, approaching path planning, and mission evaluation simulation. Case studies validate the system’s capability to forecast fire and smoke propagation, plan a water-dropping trajectory based on the fire line, optimize flight paths based on the trajectory, and simulate as well as evaluate the whole firefighting mission process. The above research comprehensively constructs the model, finishes the iterative optimization, and evaluates the water-dropping strategy by simulation. The technical path and methodological framework of studying water-dropping strategies are established. The outcomes of this study provide invaluable support for the parameter inversion design of the fixed-wing extinguisher, offering decision-making assistance to commanders and supplying training scenarios for new aviation crews.

Closing the Design-Make-Test-Analyze Loop: Interplay between Experiments and Predictions Drives PROTACs Bioavailability.

The drug development landscape is expanding to include drug modalities such as PROteolysis-TArgeting Chimeras (PROTACs) and peptides, offering possibilities for previously intractable biological targets. However, with their size and chemical nature, they diverge from established frameworks for the prediction of oral bioavailability. This evolution to larger and more complex molecules necessitates new methodologies and prediction models to continuously expand on bioavailability guidelines. We describe the high-capacity adoption of two chromatographic physicochemical assays and their application for iterative compound optimization to achieve oral bioavailability. We further describe how these data underpin the continuous refinement of internal machine learning models, which guide compound synthesis decisions in the molecular design phase. Based on data for a set of 691 PROTACs, and two project examples, we confirm a sweet spot for oral bioavailability at log D values higher than the norm for small molecules and show how experimental data and prediction models synergize to effectively drive chemistry optimization.

Clean-Label Backdoor Watermarking for Dataset Copyright Protection via Trigger Optimization

High-quality datasets are essential for training high-performance models, while the process of collection, cleaning, and labeling is costly. As a result, datasets are considered valuable intellectual property. However, when security mechanisms are symmetry-breaking, creating exploitable vulnerabilities, unauthorized use or data leakage can infringe on the copyright of dataset owners. In this study, we design a method to mount clean-label dataset watermarking based on trigger optimization, aiming to protect the copyright of the dataset from infringement. We first perform iterative optimization of the trigger based on a surrogate model, with targets class samples guiding the updates. The process ensures that the optimized triggers contain robust feature representations of the watermark target class. A watermarked dataset is obtained by embedding optimized triggers into randomly selected samples from the watermark target class. If an adversary trains a model with the watermarked dataset, our watermark will manipulate the model’s output. By observing the output of the suspect model on samples with triggers, it can be determined whether the model was trained on the watermarked dataset. The experimental results demonstrate that the proposed method exhibits high imperceptibility and strong robustness against pruning and fine-tuning attacks. Compared to existing methods, the proposed method significantly improves effectiveness at very low watermarking rates.

Nuclear norm regularized loop optimization for tensor network

We propose a loop optimization algorithm based on nuclear norm regularization for the tensor network. The key ingredient of this scheme is to introduce a rank penalty term proposed in the context of data processing. Compared to the standard variational periodic matrix product states method, this algorithm can circumvent the local minima related to short-ranged correlation in a simpler fashion. We demonstrate its performance when used as a part of the tensor network renormalization algorithms [S. Yang, Z.-C. Gu, and X.-G. Wen, ] for the critical two-dimensional Ising model. The scale invariance of the renormalized tensors is attained with higher accuracy while the higher parts of the scaling dimension spectrum are obtained in a more stable fashion. Published by the American Physical Society 2024

Efficient structural optimization under transient impact loads using multilayer perceptron and genetic algorithms

Currently, there is a lack of structural optimization methods when a structure is suffering transient impact load, considering nonlinear factors such as material plasticity and strain rate effect. The utilization of derivative-free algorithms also increases computational expenses. Accordingly, this paper presents a novel structure transient optimization approach based on multilayer perceptron combined with a derivative-free optimization method, which can efficiently solve the transient structure dynamics optimization problem with multiple optimization objects. In this approach, the area to be optimized is cut by several closed B-spline curves whose shapes are controlled by several parameters and can be varied and merged at each optimization iteration step. The structure's transient analysis process is replaced by a machine learning based surrogate model, which is trained by thousands of results from the FEM explicit transient simulation. Additionally, nearly orthogonal Latin hypercube sampling is utilized to simplify parameter dimensionality, reduce the training data set, save calculation time, and give the training data set a more comprehensive range of design parameters. In our optimization process, our design target is to minimize the peak reaction force while with the constrain of ensuring enough stiffness and mass. The results demonstrate our proposed methods could efficiently handle transient optimization problems without sensitivity calculations, exhibiting strong generalization capabilities.

Exploring multi-deformation mechanism and control of arc thin-walled structures during supercritical CO2 assisted micro milling

Titanium alloys are extensively utilized due to their exceptional corrosion resistance, high specific strength, temperature resilience, and biocompatibility. However, the challenges such as poor thermal conductivity, pronounced micro-machining size effects, and the sensitivity of micro-thin wall structures to residual stress complicate the machining of titanium alloy micro-thin walls. This paper investigates the use of supercritical CO2 to assist in arc micro-thin wall milling experiments of titanium alloys, aiming to elucidate the influence of various process parameters on micro-milling performance. The mesoscale prediction model developed in this study shows that the time-varying and static deflection deformations of micro-thin walls cooled by supercritical CO2 are approximately half of those observed under dry cutting conditions. To compare and optimize micro-milling performance metrics, an RVEA-entropy weight TOPSIS optimization scheme was developed, and combined with several high-dimensional multi-objective optimization algorithms. Integrating this with micro-milling finite element model, an iterative optimization and reverification strategy was proposed. The optimized parameters combination achieved through this method reduced micro-milling force, top deformation, and side deformation by 32.5 %, 24.6 %, and 24.3 %, respectively. The research approach and optimization strategy presented in this paper offer valuable insights for enhancing the machining precision of mesoscale titanium alloy micro-thin-wall structures.

Iterative Linear Quadratic Optimization for Nonlinear Control: Differentiable Programming Algorithmic Templates

Iterative Linear Quadratic Optimization for Nonlinear Control: Differentiable Programming Algorithmic Templates

A sensing system and solving method for dynamic detection of relative pose of hydraulic support group

The precise detection of the pose of hydraulic support groups is crucial for the construction of intelligent coal mines. Addressing current challenges in comprehensive position sensing of support groups, such as one-sided descriptions, missing information, and limited detection capabilities, this paper, based on the concept of digital twin, proposes a sensing system and solving method for dynamic detection of the relative pose of hydraulic support group. Initially, a relative pose detection model, based on the six-point localization principle, is proposed by combining existing sensor information. Then, a sensing system is designed, incorporating a photoelectric detection device for monitoring key position parameters and virtual sensors for determining pose information. Consequently, a method for dynamically deriving relative positions, driven by the fusion of real-virtual iterative solving, real-virtual error feedback, and iterative condition optimization, is established. Finally, a relative pose dynamic detection system for a hydraulic support prototype is constructed in a laboratory environment. Experimental results demonstrate that the sensor system and the relative position detection method designed in this paper achieve a relative position accuracy of over 95.32 % for the hydraulic support, with a dynamic reconstruction time of less than 0.2 s. This verifies the rationality of the sensor system and the feasibility and accuracy of the dynamic deduction method. To sum up, this work provides a theoretical foundation and technical support for the autonomous perception of hydraulic support group, autonomous frame adjustment between frames, and overall straightness adjustment. It also lays the groundwork for further research and the advancement of information-physical systems in the fully-mechanized mining face.

Research on online deformation monitoring of thin-walled parts driven by digital twin

Thin-walled parts are widely used in aerospace. Because thin-walled parts have thin walls, weak stiffness, and are easy to deform and other characteristics, it is easy to produce deformation in the processing, affecting the quality and accuracy of the workpiece processing. Therefore, accurate and efficient online real-time monitoring of the deformation of thin-walled parts has become a vital issue in process control. The visualization, real-time monitoring, and iterative optimization of digital twin technology can effectively solve the problem of online monitoring of thin-walled parts processing. Therefore, this paper proposes a digital twin driven deformation monitoring method for thin-walled parts, which realizes the effective monitoring of deformation during the processing of thin-walled parts. Firstly, the digital twin architecture for online deformation monitoring of thin-walled parts is established, and the critical technologies of the digital twin for online deformation monitoring of thin-walled parts are proposed. Secondly, the Support Vector Regression (SVR) finite element surrogate model is established to realize the rapid simulation of thin-walled parts cutting. At the same time, a One-Dimensional Convolutional Neural Network (1DCNN) combined with the Self-Attention Mechanism (SAM) deformation prediction model was established based on sensor data. The real-time monitoring and accurate prediction of the deformation of thin-walled components are realized through the parallel guidance of the deep learning and surrogate model. Finally, an experimental analysis was carried out by machining thin-walled plates of aluminum alloy 7075, and the fit of the monitoring models was above 95%, thus verifying the validity of the proposed method.