Interplay between shape and composition in bimetallic nanoparticles revealed by an efficient optimal-exchange optimization algorithm

Despite the large relevance of bimetallic metal nanoparticles for heterogeneous catalysis, the relation between their shape and elemental composition remains elusive. Here, we investigate this relationship by implementing and applying global optimization methods enhanced with a novel optimal-exchange algorithm. In particular, we determine the lowest energy chemical orderings for PtAu nanoparticles, revealing that the most stable shape changes from highly symmetric structures for pure particles to distorted and less symmetric shapes for intermediate compositions. The presented method leverages the local atomic contributions to an empirical surrogate energy expression to identify optimal atom exchanges. This also allows us to pinpoint the origin of the stability of distorted shapes, revealing a favorable energy trade-off when over-coordinating Pt and under-coordinating Au upon distorting the particle shape.

Smoothness preserving layout for dynamic labels by hybrid optimization

Stable label movement and smooth label trajectory are critical for effective information understanding. Sudden label changes cannot be avoided by whatever forced directed methods due to the unreliability of resultant force or global optimization methods due to the complex trade-off on the different aspects. To solve this problem, we proposed a hybrid optimization method by taking advantages of the merits of both approaches. We first detect the spatial-temporal intersection regions from whole trajectories of the features, and initialize the layout by optimization in decreasing order by the number of the involved features. The label movements between the spatial-temporal intersection regions are determined by force directed methods. To cope with some features with high speed relative to neighbors, we introduced a force from future, called temporal force, so that the labels of related features can elude ahead of time and retain smooth movements. We also proposed a strategy by optimizing the label layout to predict the trajectories of features so that such global optimization method can be applied to streaming data.

Global optimization with deep-learning-based acceleration surrogate for large-scale seismic acoustic-impedance inversion

Seismic acoustic-impedance (AI) inversion, which estimates the AI of the reservoir from seismic and other geophysical data, is a type of nonlinear inverse problem that faces the local minima issue during optimization. Without requiring an accurate initial model, global optimization methods have the ability to jump out of local minima and search for the optimal global solution. However, the low-efficiency nature of global optimization methods hinders their practical applications, especially in large-scale AI inversion problems (AI inversion with a large number of traces). We propose a new intelligent seismic AI inversion method based on global optimization and deep learning. In this method, global optimization is used to generate datasets for training a deep learning network and it is used to first accelerate and then surrogate global optimization. In other words, for large-scale seismic AI inversion, global optimization only inverts the AI model for a few traces, and the AI models of most traces are obtained by deep learning. The deep learning architecture that we used to map from seismic trace to its corresponding AI model is established based on U-Net. Because the time-consuming global optimization inversion procedure can be avoided for most traces, this method has a significant advantage over conventional global optimization methods in efficiency. To verify the effectiveness of the proposed method, we compare its performance with the conventional global optimization method on 3D synthetic and field data examples. Compared with the conventional method, the proposed method only needs about one-tenth of the computation time to build AI models with better accuracy.

Estimation of the numerical efficiency of global optimization methods

Currently, test problems are used to test the effectiveness of new global optimization methods. In this article, we analyze test global optimization problems to test the numerical efficiency of methods for their solution. At present, about 200 test problems of unconditional optimization and more than 1000 problems of conditional optimization have been developed. We can find these test problems on the Internet. However, most of these test problems are not informative for testing the effectiveness of global optimization methods. The solution of test problems of conditional optimization, as a rule, has trivial solutions. This allows the parameters of the algorithms to be tuned before these solutions are obtained. In test problems of conditional optimization, the accuracy of the fulfillment of constraints is important. Often, small errors in the constraints lead to a significant change in the value of an objective function. Construction of a new package of test problems to test the numerical efficiency of global optimization methods and compare the exact quadratic regularization method with existing methods.The author suggests limiting oneself to test problems of unconstrained optimization with unknown solutions. A package of test problems of unconstrained optimization is pro-posed, which includes known test problems with unknown solutions and modifications of some test problems proposed by the author. We also propose to include in this package J. Nie polynomial functions with unknown solutions. This package of test problems will simplify the verification of the numerical effectiveness of methods. The more effective methods will be those that provide the best solutions. The paper compares existing global optimization methods with the exact quadratic regularization method proposed by the author. This method has shown the best results in solving most of the test problems. This paper presents some of the results of the author's numerical experiments. In particular, the best solutions were obtained for test problems with unknown solutions. This method allows solving multimodal problems of large dimensions and only a local search program is required for its implementation.

Antenna Resonant Frequency Modeling based on AdaBoost Gaussian Process Ensemble

The design of electromagnetic components generally relies on simulation of full-wave electromagnetic field software exploiting global optimization methods. The main problem of the method is time consuming. Aiming at solving the problem, this study proposes a regression surrogate model based on AdaBoost Gaussian process (GP) ensemble (AGPE). In this method, the GP is used as the weak model, and the AdaBoost algorithm is introduced as the ensemble framework to integrate the weak models, and the strong learner will eventually be used as a surrogate model. Numerical simulation experiment is used to verify the effectiveness of the model, the mean relative error (MRE) of the three classical benchmark functions decreases, respectively, from 0.0585, 0.0528, 0.0241 to 0.0143, 0.0265, 0.0116, and then the method is used to model the resonance frequency of rectangular microstrip antenna (MSA) and coplanar waveguide butterfly MSA. The MRE of test samples based on the APGE are 0.0069, 0.0008 respectively, and the MRE of a single GP are 0.0191, 0.0023 respectively. The results show that, compared with a single GP regression model, the proposed AGPE method works better. In addition, in the modeling experiment of resonant frequency of rectangular MSA, the results obtained by AGPE are compared with those obtained by using neural network (NN). The results show that the proposed method is more effective.