A hybrid-model optimization algorithm based on the Gaussian process and particle swarm optimization for mixed-variable CNN hyperparameter automatic search

Convolutional neural networks(CNNs)have been developed quickly in many real-world fields.However,CNN’s performance depends heavily on its hyperparameters,while finding suitable hyperparameters for CNNs working in application fields is challenging for three reasons:(1)the problem of mixed-variable encoding for different types of hyperparameters in CNNs,(2)expensive computational costs in evaluating candidate hyperparameter configuration,and(3)the problem of ensuring convergence rates and model performance during hyperparameter search.To overcome these problems and challenges,a hybrid-model optimization algorithm is proposed in this paper to search suitable hyperparameter configurations automatically based on the Gaussian process and particle swarm optimization(GPPSO)algorithm.First,a new encoding method is designed to efficiently deal with the CNN hyperparameter mixed-variable problem.Second,a hybrid-surrogate-assisted model is proposed to reduce the high cost of evaluating candidate hyperparameter configurations.Third,a novel activation function is suggested to improve the model performance and ensure the convergence rate.Intensive experiments are performed on image-classification benchmark datasets to demonstrate the superior performance of GPPSO over state-of-the-art methods.Moreover,a case study on metal fracture diagnosis is carried out to evaluate the GPPSO algorithm performance in practical applications.Experimental results demonstrate the effectiveness and efficiency of GPPSO,achieving accuracy of 95.26%and 76.36%only through 0.04 and 1.70 GPU days on the CIFAR-10 and CIFAR-100 datasets,respectively.

Data Centric Design:A New Approach to Design of Microstructural Material Systems

Building processing,structure,and property(PSP)relations for computational materials design is at the heart of the Materials Genome Initiative in the era of high-throughput computational materials science.Recent technological advancements in data acquisition and storage,microstructure characterization and reconstruction(MCR),machine learning(ML),materials modeling and simulation,data processing,manufacturing,and experimentation have significantly advanced researchers’abilities in building PSP relations and inverse material design.In this article,we examine these advancements from the perspective of design research.In particular,we introduce a data-centric approach whose fundamental aspects fall into three categories:design representation,design evaluation,and design synthesis.Developments in each of these aspects are guided by and benefit from domain knowledge.Hence,for each aspect,we present a wide range of computational methods whose integration realizes data-centric materials discovery and design.

Dynamic Response Solution of Multi-Layered Pavement Structure Under FWD Load Appling the Precise Integration Algorithm

The pavement layered structures are composed of surface layer,road base and multi-layered soil foundation.They can be undermined over time by repeated vehicle loads.In this study,a hybrid numerical method which can evaluate the displacement responses of pavement structures under dynamic falling weight deflectometer(FWD)loads.The proposed method consists of two parts:(a)the dynamic stiffness matrices of the points at the surface in the frequency domain which is based on the domain-transformation and dual vector form equation,and(b)interpolates the dynamic stiffness matrices by a continues rational function of frequency.The mixed variables formulation(MVF)can treat multiple degree of freedom systems with considering the coupling term between degree of freedoms.The accuracy of the developed method has been demonstrated by comparison between the proposed method and published results from the other method.Then the proposed method can be applied as a forward calculation technique to emulate the falling weight deflectometer test for multi-layered pavement structures.