The Algorithm Study Of Support Vector Machine In Flexible Tactile Sensing Array Based On Linear Optimization

Linear programming(LP)is regarded as the most common optimization problem in daily life.Linear programming is applied in a wide range of applications such as military combat,financial wealth systems,business management analysis,and currently the latest machine learning field of statistical optimization calculations.Linear programming has a long history of hundreds of years,and the best method for solving linear programming is the simplex method,which is popular because it is efficient and simple to use.However,it lacks a rigorous theory to testify the time complexity,and there are special cases that demonstrate the limitations of the simplex method,which leads to an exponential presentation of time complexity.Afterwards,other algorithms,such as interior point method and ellipsoidal method,have been studied,which can effectively guarantee the time complexity,but they still have limitations due to the harsh operational conditions,complicated transformation process,and not strong polynomial time algorithms.So how to find the strong polynomial time algorithm on the thought of simplex method is also a crucial issue in this field of research.Support Vector Machine(SVM),as one of the best applications of statistics,is known for its efficiency,accuracy,high degree of fitting as well as good generalization,and the traditional support vector machine is also based on the Euclidean measure of distance,usually based on 2-norm Euclidean distance and mahalanobis distance.2-norm will produce better physical interpretation,but the corresponding 2-norm will also produce complex higher-order optimization problems,which will be much slower in the high-dimensional training.So how to improve the computational speed is also an important research direction of support vector machines.Flexible sensing materials are also the current frontier research field in materials science.With the constant development of computer technology,and the gradual enhancement of artificial intelligence algorithms,the current requirements for hardware are also increasingly high,and flexible tactile materials with its softness and stretchability,can be used as research carriers in electronic skin and other cutting-edge medical as well as engineering fields.The preparation process of flexible materials is also a difficult problem in materials science,the deployment of artificial intelligence-related algorithms on new materials is also the key innovative section of this paper.The main research of this paper is as follows:1.Introduce the traditional simplex method and its geometric meaning,by means of the analysis of the simplex method,the theory of prismatic cutting is proposed to provide another intuitive explanation for the rotation of the simplex method,and the stopping boundary of the optimization problem is defined by linear adjustment planning.When the blocking surface is encountered in the prismatic cutting,the projection problem of the sliding gradient algorithm on the constrained surface needs to be used to change the trajectory.In regard of the projection problem,the paper also proposes a hat matrix to make the projection calculation easier and faster,and introduces a new method of operation process for a linear programming problem through a practical example.2.Introduce the machine theory,statistical theory and also the support vector machine concept resulting from them,which can be used to solve different problems through the norm selection of support vector in support vector machine.Traditional support vector machines use the 2-norm issue,which corresponds to optimizing a quadratic programming issue,a special case of a nonlinear programming issue,while using1-norm and infinite norm can make the support vector machine kernel into a linear programming issue solved by using the sliding gradient algorithm in Chapter 2 for the linear programming issue,the L₁GSSVM and L?GSSVM algorithms are proposed to solve the classification issue,and experiments are conducted on the UCI database.The results can obtain similar accuracy as the traditional SVM algorithm,and can be more advantageous in term of time.The average reduction in CPU training time is close to 15?20%on small dimensional datasets and close to 30?35%on large dimensional datasets,and the accuracy rate has been improved on large dimensional datasets to some extent.3.The application area of flexible tactile sensing arrays is introduced,and an innovative method is proposed to dissolve thermoplastic polyurethane elastomer rubber(TPU)in dimethylformamide(DMF)at 30%solid content,and brush liquid metal electrodes hard onto smooth acrylic plates using screen printing,then brush the electrodes hard onto TPU films using a transfer technique.By using this technique,it is possible to obtain a more outstanding capacitance-pressure linear correlation ability of the new material in the characterization analysis,and better performance of adhesion and stretchability.This material could be used in the robotic arm,and the system architecture was designed to finish the receiving and sending data circuit module,the pre-processing module of data,as well as the data storage module and the most critical gesture training module,which was trained with the gradient descent algorithm of multilayer neural network in deep learning to obtain the gesture recognition capability of the robotic arm.By comparing the linear SVM algorithm with and without adding gesture parameters,the average accuracy rate can be improved by 15%after adding gesture parameters,and the training speed can be improved by 50%with L₁GSSVM and L?GSSVM algorithms.