Scale adaptive fitness evaluation‐based particle swarm optimisation for hyperparameter and architecture optimisation in neural networks and deep learning

Research into automatically searching for an optimal neural network(NN)by optimi-sation algorithms is a significant research topic in deep learning and artificial intelligence.However,this is still challenging due to two issues:Both the hyperparameter and ar-chitecture should be optimised and the optimisation process is computationally expen-sive.To tackle these two issues,this paper focusses on solving the hyperparameter and architecture optimization problem for the NN and proposes a novel light‐weight scale‐adaptive fitness evaluation‐based particle swarm optimisation(SAFE‐PSO)approach.Firstly,the SAFE‐PSO algorithm considers the hyperparameters and architectures together in the optimisation problem and therefore can find their optimal combination for the globally best NN.Secondly,the computational cost can be reduced by using multi‐scale accuracy evaluation methods to evaluate candidates.Thirdly,a stagnation‐based switch strategy is proposed to adaptively switch different evaluation methods to better balance the search performance and computational cost.The SAFE‐PSO algorithm is tested on two widely used datasets:The 10‐category(i.e.,CIFAR10)and the 100−cate-gory(i.e.,CIFAR100).The experimental results show that SAFE‐PSO is very effective and efficient,which can not only find a promising NN automatically but also find a better NN than compared algorithms at the same computational cost.

Adaptive Fitting Capacity Prediction Method for Lithium‑Ion Batteries

Lithium-ion batteries have become the mainstream power source for electric vehicles because of their excellent performance.However,lithium-ion batteries still experience aging and capacity attenuation during usage.It is therefore critical to accu-rately predict battery remaining capacity for increasing battery safety and prolonging battery life.This paper first adopts the metabolism grey algorithm and a simplified electrochemical model to predict battery capacity under different operating conditions.To improve the prediction performance where the capacity changes nonlinearly,a decoupling analysis of battery capacity loss is then conducted based on the simplified electrochemical model.Finally,an adaptive fitting method is devel-oped for capacity prediction,aiming at improving the prediction accuracy at the inflection point of battery capacity diving.The prediction results indicate that the developed adaptive fitting method can achieve high prediction accuracy under battery capacity attenuation at different discharge stages with errors lower than 2.2%.And the battery capacity decay shows linear variation,and the proposed method effectively forecast the inflection point of battery capacity diving.

Fish-Eye Image Distortion Correction Based on Adaptive Partition Fitting

The acquisition of images with a fish-eye lens can cause serious image distortion because of the short focal length of the lens.As a result,it is difficult to use the obtained image information.To make use of the effective information in the image,these distorted imagesmust first be corrected into the perspective of projection images in accordance with the human eye’s observation abilities.To solve this problem,this study presents an adaptive classification fitting method for fish-eye image correction.The degree of distortion in the image is represented by the difference value of the distances fromthe distorted point and undistorted point to the center of the image.The target points selected in the image are classified by the difference value.In the areas classified by different distortion differences,different parameter curves were used for fitting and correction.The algorithm was verified through experiments.The results showed that this method has a substantial correction effect on fish-eye images taken by different fish-eye lenses.

A Realization Method for Transforming a Topology Optimization Design into Additive Manufacturing Structures

Topology optimization is a powerful design approach that is used to determine the optimal topology in order to obtain the desired functional performance. It has been widely used to improve structural performance in engineering fields such as in the aerospace and automobile industries. However, some gaps still exist between topology optimization and engineering application, which significantly hinder the applica- tion of topology optimization. One of these gaps is how to interpret topology results, especially those obtained using the density framework, into parametric computer-aided design （CAD） models that are ready for subsequent shape optimization and manufacturing. In this paper, a new method for interpreting topology optimization results into stereolithography （STL） models and parametric CAD models is pro- posed. First, we extract the skeleton of the topology optimization result in order to ensure shape preser- vation and use a filtering method to ensure characteristics preservation. After this process, the distribution of the nodes in the boundary of the topology optimization result is denser, which will benefit the subsequent curve fitting. Using the curvature and the derivative of curvature of the uniform B-spline curve, an adaptive B-spline fitting method is proposed in order to obtain a parametric CAD model with the fewest control points meeting the requirement of the fitting error. A case study is presented to pro- vide a detailed description of the proposed method, and two more examples are shown to demonstrate the validity and versatility of the proposed method.

Modeling and Solving Human Arm’s Posture in Reachablity Analysis

Reachability is a key criterion in maintenance design, and human arm is the main object in reachability analysis. The human arm＇s DOF is reduced, and applying military standards and human physiological constraints, the simplified arm model of 7-DOF using D-H method is built up. Particle Swarm Optimization （PSO） is used to acquire the shoulder, arm and hand posture with adaptive fitness function. A detailed reachability analysis is accomplished for disassembling the bolts from crank shaft is given as an example to validate the feasibility of using teachability analysis on maintenance design.

Accuracy assessment and error analysis for diameter at breast height measurement of trees obtained using a novel backpack LiDAR system

Background: The LiBackpack is a recently developed backpack light detection and ranging(LiDAR) system that combines the flexibility of human walking with the nearby measurement in all directions to provide a novel and efficient approach to LiDAR remote sensing, especially useful for forest structure inventory. However, the measurement accuracy and error sources have not been systematically explored for this system.Method: In this study, we used the LiBackpack D-50 system to measure the diameter at breast height(DBH) for a Pinus sylvestris tree population in the Saihanba National Forest Park of China, and estimated the accuracy of LiBackpack measurements of DBH based on comparisons with manually measured DBH values in the field. We determined the optimal vertical slice thickness of the point cloud sample for achieving the most stable and accurate LiBackpack measurements of DBH for this tree species, and explored the effects of different factors on the measurement error.Result: 1) A vertical thickness of 30 cm for the point cloud sample slice provided the highest fitting accuracy(adjusted R2= 0.89, Root Mean Squared Error(RMSE) = 20.85 mm);2) the point cloud density had a significant negative, logarithmic relationship with measurement error of DBH and it explained 35.1% of the measurement error;3) the LiBackpack measurements of DBH were generally smaller than the manually measured values, and the corresponding measurement errors increased for larger trees;and 4) by considering the effect of the point cloud density correction, a transitional model can be fitted to approximate field measured DBH using LiBackpackscanned value with satisfactory accuracy(adjusted R2= 0.920;RMSE = 14.77 mm), and decrease the predicting error by 29.2%. Our study confirmed the reliability of the novel LiBackpack system in accurate forestry inventory, set up a useful transitional model between scanning data and the traditional manual-measured data specifically for P.sylvestris, and implied the applicable substitution of this new approach for more species, with necessary parameter calibration.